e DOCUMENT RESUME ED 240008
Transcript of e DOCUMENT RESUME ED 240008
e DOCUMENT RESUME
ED 240008 CS 208 180
AUTHOR Hayes, John R.;! Flower, Linda S.TITLE , A Cognitive Model of the Writing Process in Adults.
..-..- Final Report.INSTITUTION Carnegie-Mellon Univ., Pittsburgh, Pa.SPOIL'S AGENCY National Inst. of Education (ED), Washington, DC.PUB DATE 11 Oct 83GRANT NIE-F-78-01.95NOTE 92p.; Figures and tables may be marginally
legible.PUB TYPE Reports - Research/Technical (143)
EDRS PRICE MF01/PC04 Plus Postage.DESCRIPTORS Academic Aptitude; *Cognitive Processes; College
Students; Expository Writing; Higher Education;Models; Planning; *Research Methodology; WritingImprovement; *Writing Processes; *Writing Research;*Writing Skills
IDENTIFIERS *Protocol Analysis
ABSTRACTA research project was undertaken to (1) identify the
major cognitive processes involved in expository writing, (2) test amodel of the organization of those processes, and (3) identifyteachable aids that could be used by poor and average adult writersto improve the'r writing skills. Subjects were expert and novicestudent writers at Carnegie-Mellon University in Pittsburgh,Pennsylvania. The research method employed was that of protocolanalysis, with each student being asked to think aloud as he or sheperformed writing tasks. The resulting tapes were analyzed todiscover the cognitive processes involved in writing and to develop amodel of composition. The findings suggest the following: (1) thereare important differences in the ways expert and novice writershandle the writing process, (2) many of the strategies employed byexpert writer! are teachable, and (3) one of the most promising areasfor improving students' writing is in tha art of planning. (Charts,tables of data, and excerpts from student protocols are appended.)(FL)
***********************************************************************
Reproductions supplied by EDRS are the best that can be made* from the original document.**********************************************************t************
COOO
1.1.1
U 3 OEPARTMENT OF EDUCATIONNATIONAL INSTITUTE OF EDUCATION
EDUCATIONAL RESOURCES INFORMATION
,X TN% doCENTER LEM)
onnent KIS bee. eprodoced es.P.I owq$ front the IWISOO or otganoehonotortehon 4kin°, . PW..$ Kto. been oxide to onprot eetIrOtigl boo hthent,
P. of rwor a volloor stated et this documen' alt not net eSNtosttt wprescrtt ("rt Mtirrution 40 vet r
Final Reporton Grant NIE-F-78-0195
A Cognitive Model of the Writing Process in Adults
John R. Hayes and Linda S. Flower
Carnegie-Mellon University
Pittsburgh, PA 15213
11 October 1983
4
Hayes/Flower
ABSTRACT
The ability to write effectively is important for achievemett in post-secondary education and in professional life. Yet the ability to usewriting as a practical and intellectual skill has eluded many adults.The three main objectives of this research project on adult writingwere to identify the major cognitive processes involved in expositorywriting; to test a model of the organization of these processes; and toidentify teachable aids which could be used by poor and average adultwriters to improve their writing skills.
Subjects for the project were competent and non-competent writers atCarnegie-Mellon University. The research method employed was that ofprotocol analysis; each student was asked to think aloud as he performedwriting tasks. These protocols enabled the principal investigators tobetter understand the cognitive processes involved in writing and todevelop a model of composition.
This research on the composing process has a number of implications forteaching including these three important observations: 1.) There areimportant differences in how expert and novice writers handle the processof writing; 2.) Many of the strategies employed by experienced writersare teachable; and 3.) One of the most promising areas for improvingstudents' writing is in the neglected art of planning. Teaching studentsto plan what and what to say and to learn while they write can offerthem a very useful skill.
3
nirrilibMINMAIMIlar.
i
Table of Contents1. Introduction 1
1.1. Strategic Decision:. 1
1.2. The Task Environmont 6,.3. The Writers Long-Term Memory 6
2. A Model of the Writing Process 62.1. The Writing Prccess 72.2. Planning: Generatlrg 7
2.3. Planning: Orgarizing 8
2.4. Planning: Goo: Setting 9
2.4.1. Translating 9
2.4.2. Reviewing 10
2.4.3. The Monitor 12
2.5. Testing tne Model 132.6. Protocol Sections 15
2.7. HYPOTHESIS 1: The Form of the Written Materials 16
2.8. HYPOTHESIS 2: Classifying "Content" Statements 17
2.9. HYPOTHESIS 3: Measuring Retrieval Chains 202.10. Conclusions 21
3. The Dynamics of Composing 22
3.1. Introduction 22
3.2 Knowledge 23
3.3. Written Speech 24
4 Cognition of Discovery 26
4.1. Studying Cognitive Processes 27
4 2. A Model of the Rhetorical Problem 284.3. Differencees Among Writers 28
5 A Taxonomy of Writing Plans 31
5 1. Plans To Do 31
5 2 Piens To Stsy 32
5 3 Composing Plans 32
6. The Pregnant Pause 35
6.1. Evidence for an Episode Structure 37
7. Evaluating How Writers Generate !leas 39
7 1. Results 42
8 Furmulat,ng Sentences :n Writing 43
8 1 The Relation Bet.yeen Planning and Sentence Construction 44
82 What Happens When Sentences Are Written? 47
8 3 Diffeeences Bi.:Neen Experts and Non-experts 49
9. Implications for Teaching 51
9.1 Implication 1 51
, 9.2. Implication 2 53
9.3 Implication 3 54
9 4 Teaching Planning Versus Doing it 57
9 5 A Plan To Do 59
96 A Plan To Sav 60
9 7 A Plan To Discover 61
1
1. Introduction
During the last three years of research supported by ME, we have applied protocol
analysis and other methods of cognitive science to the analysis of written composition.
In conducting our research we have made a number of strategic decisions about what is
interesting and about how best to proceed. These decisions are the incarnation of our
scientific biases.
1.1. Strategic Decisions
Our approach proceeds from five strategic decisions which we made about how to
conduct our research. Briefly these decisions were:
1. to focus on the act of writing:
2 to try for a process model of writing,
3. to model individual writers;
4. to work wholistically or lop down"; and
5. to divide the writing task ori.o parts for easier analysis.
As we will see below, these decisions are genuine ones in the sense that we could
reasonably have made other choices. Alternative approaches to the study of writing
proceed from different decisions on these same issues.
1. Our first and most important decision was to focus on the act of writing -- that
is, to attend to vvriatever it is that writers do when they produce a text. Thus, we viewed
writing primarily as a process rather than as a product We felt tnat by far the richest
source of information about writing would be to observe step by step how the writer had
actually created the essay. However, we did not intend to ignore the product. Wherever
possible, we looked to the writer's essay for evidence to confirm or elaborate the more
direct observations of process
To observe writers in action we have employed process training methods borrowed
2
from cognitive psychology. In our studies, a typical experiment proceeds as follows:
subjects appear at the experimental session knowing that they will be assigned a topic on
which to write an essay and that the whole procedure will take about an hour. Further,
they know that they will be asked to "think aloud" while writing. The subject is seated in
a quiet office with a desk pencil, and paper, and the tape recorder is turned on. The
experimenter then gives the subject an envelope containing the writing assignment
-- that is, the topic and the intended audience. The subject then busily sets to work
writing and commenting roughly as follows: "Well, open up the magic envelope. OK.
Whew' This is a killer. Write about abortion pro and con for Catholic Weekly. OK boy'
How am I going to handle this?", etc This continues for about an hour until the subject
says something like. "Weil, that's it Good bye, tape recorder (click]." The data of the
study consist of a verbatim transcript of the tape recording (with all the "um's" and
pauses and expletives undeleted) together with the essay and all of the notes the writer
has generated along the way. The transcript is called a protocol. These materials are
then examined in considerable detail for evidence which may reveal something of the
processes by which the writer has created the essay. In general, the data are very rich in
such evidence Subjects typically give many hints about their plans and goals. e g ,
just jot down ideas as they come to me, about strategies for dealing with the audience.
e g. 111 write this as if I were one of them", about criteria for editing and evaluaton, e g
"For 10-year-olds, we better keep thi$ simple", and so on. The analysis of this data is
celled protocol analysis.
2 To understand the writing act, we certainly need to identify the processes
involved -- but this is not enough. We also need to know how these processes are
organized to produce a text. That is, we need to know how the processes are sequenced.
how one process is terminated and how the one which follows is chosen. how errors are
detected. etc. Further, we want to know how simultaneous processes interact When
3
writers construct sentences, we want to know how they handle such multiple constraints
as the requirement for correct grammar, appropriate tone, accuracy of meaning, and
smooth transition. In short. we want a model which specifies the processes involved in
writing and accurately describes their organization and interaction.
A model is a metaphor for a process: it's a way to describe something, such as the
composing process, which refuses to sit still for a portrait. People build models in order
to understand how a dynamic system works. and to describe the functional relationships
among its parts. In addition, if a model is to really help us understand more. it should
speak to some of the critical questions in the field of writing and rhetoric. It should help
us see things in a way we didn't see t:.em before
Our second strategic decision was to direct our research toward the construction of
such a model Ideally. the model should be capable of telling us how writers go about
producing a text when they are given a writing assignment. It should tell us what
processes are involved, in what order they occur, and at what points the writer will
experience difficulty. At present, of course. we must be satisfied with a model which is
much less complete that the ideal The ideal defines where we would like to go. but,
alas'. not where we are now.
.3. It is apparent that not all writers write in the same way. For example, some
writers plan their essays from beginning to end before they write a single word of text,
whde others never seem to look beyond the next sentence. Further, some writers seem
to write with their rea:!es constantly in mind. checking frequently to be sure that they
have taken the ieade's knowledge and attitudes into account. Others appear serenely
unaware that an audience could fail to understand what they, in good faith, have intended
to say.
In modeling. we can deal with such differences in either of two ways. We can
choose to construct a model of the "'average" writer and delay until some more propitious
4
time the description of differences among writers. This approach has the merit of
simplicity. Further, if things work out well, a model of an average writer might be useful
in characterizing individual differences. Thus, models for individual writers might prove to
be minor variants of the average model. However, thiS approach may have the
disadvantage that averages sometimes suffer from -- the average may be representative
of no one. Thus, we sincerely hope that no one has the average number of children
-- two and a half -- nor would we want anyone to have to eat an average course at
dinner, which might be a compromise between appetizer and dessert such as oysters
with chocolate sauce.
An alternative approach is to construct models which are intended to describe
individuals rather than averages of groups. The disadvantage of this approach is that it
may be expensive In the worst case, each individual may require a separate model. With
better luck. models of individual writers will turn out to be variants of a small number of
model types. The advantage of this approach is that it is more likely than a model of the
average to capture the behavior of actual (rather than idealized) writers
Our third strategic decision, then, was to model the behavior of individual writers
rather than the average behavior of groups of writers.
4. in studying writing, we might well have started with processes which
psychologists and psycholinguistics have already identified as fundamental ones
-- processes such as short-term memory, grammatical categorization, and lexical
marking We might then have attempted to synthesize more complex processes using
these fundamental processes as building blocks, This synthetic or bottom up approach is
a very familiar one in science and has frequently been used with great success.
Geometry and Newtonian physics are perhaps the best known examples
However, research often proceeds in the opposite direction, that is, wholistically, or
from the top down Chemistry provides a good example of top down research. Chemical
5
research often starts with a complex compound and then looks for the elementary
components and their relations. The top down approach is the one we have chosen to
apply in our writing research. We have started from the top with the complete writing
act and have attempted to analyze it first into a few relatively complex subprocesses. As
the analysis proceeds, the complex subprocesses are analyzed further into progressively
simpler subprocesses Ultimately, we hope that this top down analysis will make contact
with the fundamental processes which psychologists and psycholinguistics have already
identified. Thus, the top down and bottom up approaches may be viewed as
complementary.
The advantage of the bottom up approach is that it is rooted in fundamental
processes. The advantage of the top down approach is that its results are almost certain
to be relevant to real writing situations.
5. Our final strategic decision was to divide the writing task into three parts (see
Figure 2 1):
1. The task environment -- that ;s, the world outside the writer's skin.
2 The writer's long-term memory; and
3 The writing processes that is, the writer excluding the writer's long-termmemory.
We chose this division because it is an especially convenient one for psychological
ana!ysis and rnode!Iing. Transfers of information between the task environment and the
writer are usud:ly marked clearly by overt acts of reading or writing Further, information
retrie:al from long-term memory is frequently detectable by examining the verbal
protocol Thus, the boundaries we have chosen divide the writing task into parts whose
interactions are relatively easy to observe
Bitzer's analysis of the rhetorical situation (1968) focuses on the importance of the
task environment Lowes' classic study of Coleridge (1927) focuses on the importance of
0
6
the writer's long-term memory. Our own research has focused on the writing processes
1.2. The Task Environment
The task environment includes everything outside the writer's skin that influences
the performance of the task_ It includes the writing assignment, that is, a description of
the topic aria the intended audience, and it may include information relevant to the
writer's motivation. For example, the teac;ier's stern expression when he presents an
assignment may tel: the writer that the assignment must be taken very seriously. Britton
et al. (1975) have emphasized the importance of such motivational factors. Once writing
has begun, the task environment also includes the text which the writer has produced so
far. This text Is a very important part of the task environment because the writer refers
to it repeatedly during the process of composition.
1.3. The Writer's Long-Term Memory
We assume that writers have knowledge about many topics, e.g., auto mechanics
and American history, and about many audiences, e g., children and Catholics, stored in
long-term memory. They may also have generalized writing plans, perhaps in the form of
a story grammar (Rumelhart, 1975) or a formula such as the journalist's questions, ''who,
what, where, when, why?".
2. A Model of the Writing Process
The unique features of the mod& are
1. It identifies not only subprocesses of the composing process, but also theorganization of those subprocesses
2. Minor variations in its simple control structure (shown in Figure 2.6) allow itto describe individual differences in composing styles
Although the model is provisional, it provides a first approximate description of
normal composition that can guide research and afford a valuable starting point in the
7
search for more refined models.
2.1. The Writing Process
We propose that writing consists of three major processes: PLANNING,
TRANSLATING, and REVIEWING. The PLANNING process consists of GENERATING,
ORGANIZING, and GOAL-SETTING subprocesses. The function of the PLANNIN-G process is
to take information from the task environment and from long-term memory and to use it
to set goals and to establish a writing plan to guide the production of a text that will
meet those goals. The plan may be drawn in part from long-term memory or may he
formed anew within the PLANNING process The TRANSLATING process acts under the
guidance of the Kitting plan tLi produce language corresponding to information in the
writer's memory. The function of the REVIEWING process. which consists of READING and
EDITING' subprocesses, is to improve the quality of the text produced by the
TRANSLATING process. It ooes this by detecting and correcting weaknesses in the text
with respect to language conventions and accuracy of meaning. and by evaluating the
extent to which the text accomplishes the writer's goals. The structures of the various
processes are shown in Figures 22 through 2.6.
2.2. Planning: Generating
The function of the GENERATING process is to retrieve information relevant to the
vaiting task from long-term memory We assume that this process derives its first
memory probe frorr information about the topic and the audience presented in the task
environment Because each retrieved item is used as the new memory probe. items are
retrieved from memory in associative chains In order to focus search on relevant
material, the retrieval chain is broken whene;:er an item is retrieved that is not useful to
the writing task Search is then restarted with a new memory probe derived from the
task environment or from useful material already retrieved
Some criterion for terminating search chains is essential to prevent the process
from getting lost in associative reverie. The cr terion that we have chosen, i.e., one
irrelevant item, may have to be relaxed somewhat to simulate human performance
accurately. We believe, though, that it won't have to be rehmed much beyond one item.
The most persistent memory searches we have observed in writing protocols never
extended more than three retrievals beyond useful material.
Insert Figures 21 and 2.2 about here
When an 'item is retrieved, the GENERATING process may produce a note.
Characteristically, these notes are single words or sentence fragments, although they may
sometimes be complete sentences The form of these notes will be used later to identify
occurrences of the GENERATING process
2.3. Planning: Organizing
The function of the ORGANIZING process is to select the most useful of the
materials retrieved by the GENERATING process and to organize them into a writing plan.
The plan may be structured either temporally (e g, "First, I'll say A, then B.") or
hierarchically (e g., "Under topic #1, I should discuss A, 8, and C.") or both.
Organizing is done by the elementary operators shown in Figure 23 The first four
of these operators act on single topics, or pairs of topics, e g, the second operator
dec,des which of two topics to discuss fif!t The :ast operator. "Identify a category," may
act to classify a large number of topics that were generated separately under the same
, heading
insert Figure 2.3 about here
Notes generated by the ORGANIZING process often have an oiganizational form
12 1
9
That is, they are systematically indented, or numbered, or alphabetized, or possibly all of
these. This organizational form will be used later to identify occurrences of the
ORGANIZING process.
2.4. Planning: Goal Setting
Some of the materials retrieved by the GENERATING process are not topics to be
written about but rather are criteria by which to Judge the text. Often such criteria
appear in the protocol when the writer is considering the audience or features of the text.
At such times the writer ma., say, "Better keep it simple," or, 1 need to write a transition
here - The GOAL SETTING process identifies and stores such criteria for later use in
EDITING.
2.4 1. Translating
The function of the TRANSLATING process is to take material from memory under
the guidance of the writing plan and to transform it into acceptable written English
sentences. We assume that material in memory is stored as propositions but not
necessarily as language By a proposition, we understand a structure such as
[(Concept A) (Relation B) (Concept C))Or
[(Concept 0) (Attribute E)], etc.
where concepts, relations, and attributes are memory structures, perhaps complex
networks or images, for which the writer may or nay not have names
To illustrate the operation of the TRANSLATING process (see Figure 2 0), we have
invented a scenario of a student writing an essay on Henri Rousseau.
1. Get next part of writing plan. "I've covered the early years, not' I'vegot to say how he got into painting."
2. Plan next sentence: Retrieve propositionsProposition A: [(Rousseau) (showed) (sole early promise))Proposition B' ((Rousseau) (did) (very little painting until 40)1Sentence plan: (Pi.:position A) but (Proposition B)
13....111.
i
10
3. Express next proposition part: "Rousseau ... Rousseau, what? Rousseaudisplayed ... Although Rousseau displayed some early prt - ise ..., etc.'
Insert Figure 2.4 about here
Writing done during the TRANSLATING process shows two features:
1. Characteristically, it is in the form of complete sentences, and
2. It is often associated with a protocol segment that contains an interrogativereflecting search for the next sentence part, e.g., "Rousseau did what ?" or,"How do I want to put this?"
These features will be used later to identify occurrences of the TRANSLATING
process.
2.4.2. Reviewing
The function of the reviewing process is to improve the quality of the written text.
It consists, as Figure 2.5 shows, of two subprocesses: READING and EDITING.
Insert Figure 2.5 about here
Reviewing: Edit;ng. The EDITING process examines any material that the writer
puts into words, whether by reading, writing, or speaking. Its purpose is to detect and
correct violations in writing conventions and inaccuracies of meaning and to evaluate
materials with respect to the writing goals. These evaluations may be reflected in
4_,,Jestions such as. "Will this argument be convincing?" and. "'Jaye I covered all parts of
the plan?"
11
We assume that the EDITING process has the form of a production system.1 The
conditions of the productions have two parts. The first part specifies the kind of
language to which the editing production applies, e.g formal sentences, notes, etc. The
second is a fault detector for such problems as grammatical errors, incorrect words, and
missing context. When the conditions of a production are met, e.g., a grammatical error
is found in a formal sentence, the action that is triggered is a procedure for fixing the
fault.
Consider the following production:
[(formal sentence) (first letter of sentence lower case)> change first letter to upper case]
If the writer is producing formal sentences, this production will detect and correct errors
in initial capitalization. However, if the writer is only producing notes, the conditions of
the production will not be met and capitalization will be ignored.
Although the action in the preceding production is simple, in some cases the action
may invoke the whole writing process recursively. For example, in one writing protocol,
the w; tees first draft contained the first sentence of tt final draft immediately followed
1A production system is an ordered sequence of condition-action rules The left sideof each rule shows the condition or stimulus, and the right side shows the action to betaken if the condition is met The conditions are tested in order, starting with the firstrule The order of the productions is important Consider the production system forputting a horse in a barn
Conditions(horse out of barn) and(barn door closed)
(horse Out of barn)(barn door open)
Actions--> (open barn door)
--> (put horse in barn)--> (close barn door)
Changing the order of these productions could have very serious consequences for thehorse!
( 15
12
by the seventh sentence ot.the final draft. In editing the first draft, the writer recognized
that the reader would not have sufficient context to understand the relation between
these two sentences. To correct this fault, the writer constructed a small explanatory
essay to insert between the sentences. Thus, in this case, the fixing procedure invoked
the whole writing process.
We assume that the EDITING process is triggere(1 automatically whenever ther
conditions of an editing production are satisfied and that it will interrupt any other
ongoing process
Insert Figure 2.6 about here
We distinguish between REVIEWING and EDITING as two distinct modes of behavior.
On the one hand, EDITING is triggered automatically and mai occur in brief episodes
interrupting other processes REVIEWING, on the other hand, is not a spur-,f-the-
moment activity but rather one in which the writer decides to devote a period of time to
systematic examination and improvement of the text. It occurs typically when the writer
has finished a translation process rather than as an interruption to that process
2.4.3. The Monitor
The relations among the processes are defined by the simple production system
shown in Figure 27. The structure of the monitor was chosen to reflect three
observations about composition processes
Insert Figure 27 about here
1. The EDITING and GENERATING processes may interrupt other processes.Thus. the first two production rules triggering EDITING and GENERATINGprocesses take priority over goal setting rules
2 The writer's intuitions and the persistence of his or her actions suggest that
16
13
writing processes are controlled by goals. Thus, if writers report that they aretrying to organize material, they will persistently return to ORGANIZINGprocesses even when those processes are interrupted by EDITING andGENERATING (productions 3 through 6 define the writer's goals).
3. Individual differences In goal setting reflect important individual differences inwriting style. Figure 2.8 shows four alternat'ae configurations for the goalsetting productions. Each configuration corresponds to a characteristicallydifferent way of producing an essay. Configuration 1, for example,corresponds to a style in which the writer tries to produce a perfect firstsentence and then to follow the perfect first sentence with a perfect secondsentence and so on. The work of planning, translating, and reviewing eachsentence is completed before the writer proceeds to the next sentence. WithConfiguration 2, thoughts are written down as they occur to the writer and hereviews them ester. With Configuration 3, the writer tries to generate aperfect first draft. Configuration 4 yields a breadth-tirst composing process.A draft is planned and then written out in fun before any review takes place.Lowenthal and Wason (1977) have described writing styles among academics
:that correspond to Configurations 3 and 4.
Insert Figure 2 8 about here
Rules 7 through 10 in Figure 2 7 have the effect of executing the current goal when
the goal activity is not being interrupted by rule 1 or 'tile 2.
As a final observation about the model, notice that the GENERATING process
operates differently when the goal is GENERATING than when it is not. When the goal is
GENERATING, the GENERATING process is persistent. That is, each attempt to generate is
followed by another attempt to generate. When the goat is not GENERATING, each
attempt to generate is followed by a return to the process specified by the current goal
(the one which GENERATING interrupted).
2.5, Testing the Model
We compare our model with a writing protocol in which the writer gave especially
clear indications of ongoing writing processes and of the transitions between processes
(The writer's style suggests that he sets his goals in the same way as the monitor with
Configuration 4 see cigure 2.8) This relatively unambiguous protocol provides a
14
rigorous test of the model's adequacy.
The protocol consisted of 14 pages of verbal transcript (the thinking aloud part of
the protocol), five pages of notes, and a page of completed essay. We divided the verbal
transcript into a sequence of segment), each containing a simple comment or statement.
We have analyzed the first 458 segments of the transcript, or about half of it.
The segments are of three general types:
1. Metacomments -- comments that writers make about the writing processitself, e.g.. "I'll just make a list of topics now," "I'm going to write out a draft,""Better go back and read it over,"
2. Task-oriented or "content" statements--statements that reflect theapplication of writing processes to the current task. e.g.. 'That's not the rightword" reflects an editing process. "I'll use that topic last" reflects an
organizing process, etc.
3 Interjectionssuch as "Ok, "Well, let's see." all righ:,""urnm,"ah,- etc
Consider tie sequences of segments. We 1111'11 just make a list of topics now./Energy
conserveion,/pollutionfunempinyment. The first segment is an interjection; the second, a
metacomment. and the rest are task-oriented statements. (Interjections were not analyzed
in this study.)
Writing protocols are complex, and writers are often incomplete or ambiguous when
they describe what they are doing. As a result, in analyzing a protocol. we frequently
have to make judgments about the writer's meaning. The presence of such judgments
may lead one to question the objectivity of the analyses Because we are testing our
model by comparing it to a protocol, we have to be especia:ly careful to establish the
objectivity of our analysis. To do this, we have taken the following steps
1. Whenever objective evidence was available, we used it. Thus, reading andwriting processes were identified by matching the verbal protocol word forword with the writer's notes and text (the objective evidence).
2. Whenever possible, processes were identified by using converging lines ofevidence. e.g.., the form of the written material on the one hand. and thewriter's comments about what he is doing on the other.
16
15
3. The mcst important analyses were replicated by independent judges.
2.6. Protocol Sections
The writer's metacomments suggest that the protocol can be divided quite cleanly
into three sections. In the first section, including segments 1 through 116, tne writer's
goal is to generate; in the second, including segments 117 through 270, it is to organize,
and in the third, including segments 271 through 458, it is to translate. Here are the
metacomments that led us to this conclusion:
Segment 2: "And what I'll do now is to simply jot down random thoughts.."
Segment 5: 'Topics as they occur randomly are.."
Segment 48: "Organizing nothing as yet."
Segment 69. "Other things to think about in this random search are
Segment 117: Now I think it's time to go back and read over the material
and elaborate on its organization."
Segment 161: "Now this isn't the overall organization. This is Just
the organization of a subpart.-
Segment 237: "There's an organization.-
Segment 239: "Let's try and write something."
Segment 243: "Oh, no We need more organizing
Segment 269: "I can imagine the possibility of an alternate plan..."
Segment 271: "But let's build on this plan and see what happens with it."
If these assumptions about goals are correct. it follows from the model that the
most frequent process in the first section will be GENERATING interrrupted occasionally
by EDITING, in the second, ORGANIZING interrupted by GENERATING and EDITING, and in
the third, TRANSLATING interrupted by GENERATING and EDITING. Further. we can make
three predictions about the protocol:
.15
16
1. The form of the written material should vary from section to sectioncorresponding to changes in process from section to section. Thus, in thefirst section, we expect the generating process to produce many single worn 3,detached phrases, and incomplete sentences. In the second section, ..,3expect the organizing process to produce material that is systematicallyindented, alphabetized, or numbered. In the third section, we expect thetranslating process to produce many complete sentences and some materialassociated in the verbal protocol with interrogatives suggesting search forsentence continuation.
2. The content statements in the protocol should relect the distribution ofprocesses just predicted, and
3. The generating process should be more persistent in section 1 than insections 2 and 3.
2.7. HYPOTHESIS 1: The Form of the Written Materials
To test the first hypothesis, we wanted to determine if items written during the first
section had a form consistent with the GENERATING process; items written during the
second section, with the ORGANIZING process, and the items written during the third
section, with the 1RANSLATING process. For this purpose we identified all of the items
written in the three protocol sections : 26 in the first section; 24 in the second; and 12 in
the third. An item was a word, phrase, or sentence that was identifiable in the verbal
protocol as being written during a single segment or several contiguous segments it
was, in effect, a short burst of writing.
Three raters were given the written material and verbal protocol and were asked
indepe-Aently to make the following judgments about each written item
1. Does it have good form. i e . is it a complete. grammatical sentence,
2 Is it part of a systernaticcIly indented, alphabetized, or numbered structure, i e,does it appear to be part of an 'utline or structured plan of some sort?
3. Is it associated in the verbal protocol with an interrogative suggesting searchfor sentence completion?
Table 21 shows that there was excellent agreement among the raters in making
these judgments. For each of the properties, Table 2 2 shows the proportion of items
17
written during each section that were judged to have that property. An item was scored
as having a property if two or more of the judges agreed that it did.
Insert Tables 2.1 and 2.2 about here
Items written during section 1 sometimes had good form but most usually had none
of the three properties. Items written during section 2 typically showed the second
property (indentation, etc.) but neither of the other properties. Two-thirds of the items
written during section 3 were of good form and many were associated in the protocol
with interrogatives. No items written in any other section were associated with an
interrogative. These results are quite consistent with the view that GENERATING is the
dominant process in section 1. ORGANIZING in section 2, and TRANSLATING in section 3,
and thus provide strong support for Hypothesis 1.
2.8. HYPOTHESIS 2: Classifying "Content" Statements
Our second hypothesis is that the content statements in the protocol will reflect
differences in distribution of processes in the three protocol sections. As with our first
hypothesis, we are looking for evidence that the writing processes we have postulated
turn up where they ought to, e g., GENERATING should appear prominently when the
writer says that his goal is to generate ideas, etc. !n addition, we are looking for
evidence that the EDITING ,..nd GENERATING processes interrupt the other processes as
we have postulated Again, the expected distribution of writing processes is. in the first
section, GENERATING interrupted by EDITING. in the second. ORGANIZING interrupted by
EDITING and GENERATING. and in the third. TRANSLATING interrupted by EDITING and
GENERATING
To test this hypothesis. each of the authors independently classified each segment
in two ways.
18
In classification 1, eachIstegment was judged as belonging to one of the following
four categories: (a) interjectigns, (b) metacomments; (c) content statements; and (d) a/
combination of metacomments and content statements.
In classification 2, the authors made judgments as to which of the writing processes
was most likely to have given rise to the segment. Four alternative writing processes
were considered: GENERATING, ORGANIZING, TRANSLATING. and EDITING.
Because the protocol sections were identified by examining the writer's
metacomments, we wanted to test Hypothesis 2 using only segments that were purely
content statements with no component of metacomment. Therefore, in the following
analysis, we have considered only those segments that both authors classified as pure
content statements. Out of a total of 458 segments, 170 were identified °s pure content
statements, approximately 130 as interjections, 18 previously irk tified as "reads* were
not judged. and the remainder were judged by one author or the other as being
metacomments in part or whole.
The authors agreed in attributing writing processes in 144 or 84.7% of the 170
content statements. Table 2.3 shows that, despite some differences, the authors agree
that the content statements in section 1 can be attributed mostly to GENERATING; in
section 2, to ORGANIZING: and in section 3, to TRANSLATING. They also agree that
approximately 10 to 15% of the segments in each section can be attributed to EDITING
and that approximately 10 to 15% of segments in sections 2 and 3 can be attributed to
GENERATING The most important disagreement is that one author attributes some
segments in sections 1 and 2 to TRANSLATING whereas the other does not
Insert Table 2 3 about here
Figure 2 9, which shows the processes author 2 attributed to the sequence of
metacomments and content statements, illustrates two features of the protocol:
19
1. Interruptions of other processes by EDITING and GENERATING are frequentand wiltiely distributed.
2. Even though in segment 117, the writer announced, Now it's time to go backand read over the material and elaborate on its organization," apparently hedoesn't do very much organizing until segment 153. The reason for this isthat the writer is indeed reading (10 "reads' occurred in this interval), and thereading triggered some GENERATING and EDITING interrupts.
Because we made the judgments of process in the context of the whole protocol,
one must be concerned that this context could have influenced our judgment. For
example,, we might have attributed a segment to GENERATING rather than to
TRANSLATING if the segment occurred early in the protocol.
To determine if consistent judgments of process could be made without context, we
conducted the following study. We selected 41 content statements from the protocol and
typed them on cards. The cards were then shuffled and presented for judgment
independently to two coders (not the authors). Coder 1 agreed with one of us in 67% of
judgments and Coder 2, in 77% of judgments. Most of the disagreements (16 out of 22)
involved judgments of EDITING. Many segments that the author attributed to EDITING the
coders attributed to GENERATING. EDITING may be especially difficult to identify out of
context because "edits" often present a comment on the previous segment or represent a
change in a previous segment It is difficult, for example, to see that segment 87, "I
guess all elements are tow lever indicates editing for redundancy unless one also sees
segment 86, "even low level elements of writing " If we consider only segments that the
author attributed to GENERATING ORGANIZING, or TRANSLATING, we find that both
coders agree with the author in 86% of cases. These high levels of agreement are very
encouraging and suggest that even if judgments were made without context, our
conclusions concerning Hypothesis 2 would be substantially the same Overall, then, our
results strongly support Hypothesis 2
20
Insert Figure 2.9 about here
2.9. HYPOTHESIS 3: Measuring Retrieval Chains
Our third hypothesis is that the GENERATING process will be more persistent during
section 1 of the protocol, when the goal is to generate, than during sections 2 and 3,
when it is not. To test this hypothesis, one of the authors identified all of the content
ideas generated during the protocol. (A single idea might be the topic of several protocol
segments but was nonetheless counted as one idea) A total of 48 separate ideas was
identified. The two authors then independently judged whether each idea rad been cued
by the previous idea or not. Because the authors' judgments agreed in 96% of cases, we
simply present the average of their results
In section 1, 32 ideas occurred in chains of average length 6.4, whereas in sections
2 and 3, 16 ideas occurred in chains of average length 2.0. As the model predicted, the
GENERATING process was much more persistent during the first section of the protocol
than during the second two. The fact that the average chain length in sections 2 and 3
was two rather than one as the model requires suggests that our criteria for terminating
search should be relaxed a bit.
The sequence in which ideas were retrieved in section 1 was strongly determined
by associative connections to appear in the final essay We might expect this unless, of
course, an active ORGANIZING process intervenes between GENERATING and
TRANSLATING as the mode! postulates.
Figure 2.10 shows the writer's outline for the essay as a structure of ideas in tree
form The numbers in the figure indicate the order in which the ideas were generated
Clearly, the retrieval order is very different from the outline order
24
-.
21
Insert Figure 2.10 about here
2.10. Conclusions filWe believe that the evidence provides very encouraging support for our model. All
three of the model's predictions were strongly confirmed. We should note, however, that
although these results are encouraging, they are quite limited in scope. First, although
the model was derived through informal analysis of many protocols, it has been tested
formally with only one protocol. Second, although the model is quite complex, only a few
of its properties have been tested. We have tested some properties of the major writing
processes, but we have not, for example, tested the model's predictions about individual
differences nor about the structure of the editing processes We plan to conduct much
more extensive testing of the model in the near future.
Whether or not it is supported by the data, one may still ask, is there really
anything new about the model? Haven't English teachers been talking about processes
such as planning, organizing, and editing for a long time?" Indeed English teachers have
been talking about such processes for a long time. Nonetheless, there is a great deal
that is new about the model. First, the model is rather speeVic about the nature of the..._ .
---.......-cindividual processes (see Figures 22 2.6). Second, and mbre important, the model\\specifies the organization of these processes In particular, it specifies an organization
that is goal directed and recursive, that allows for process interrupts, and that can
account for individual differences.
We should caution the reader not to interpret our model as a stage model. We are
not saying that writing proceeds in order through successive stages of PLANNING,
TRANSLATING, and REVIEWING. It may do so, and, indeed, in the part of the protocol
examined in this paper, writing did proceed generally in successive stages. However, this
25
4
22
is not the only sort of writing behavior we have cbserved, nor is it the only sort allowed
by the model. The model is recursive and allows for a complex intermixing of stages. As
we noted previously, the whole writing process, including PLANNING. TRANSLATING, and
REVIEWING, may appear as a part of an EDITING subprocess. Because EDITING can
interrupt any other process, these processes can appear within any other process.
Further, we should note that we do not intend to imply that all writers use all of the
processes we have described. Our model is a model of competent writers. Some
writers, though, perhaps to their disadvantage, may fail to use some of the processes.
We have, for example, observed a writer who failed to organize. This writer, however,
could not be viewed as competent
We believe that our model, if it is approximately correct, can serve as a guide to the
diagnosis of writing difficulties. We hope that, whether it is right or wrong, it can serve
as "a target to shoot at." and hence a guide to further research on writing.
3. The Dynamics of Composing
3.1. Introduction
In this section we attempt to use our proposed model of the writing process to
describe writers in action In other words, we would like to account, from the writer's
point of view, for the dy_narnics of composing. We make two major points. The first is
that the act of writing is best described as the act of juggling a number of simultaneous
constraints This is in contrast to seeing it as a series of discrete stages or steps that
add up to a finished product Second, we suggest that one of the most effective
strategies for handling this large number of constraints is Planning. Plans allow writers
to reduce "cognitive strain," that is, to reduce the number of demands being made on
conscious attention. (They also create a nested set of goals that allow a number of
constraints to be satisfied at once.)
23
In general, the constraints an adult writer must shoulder seem to fall into three
major groups of increasing inclusiveness: the second is the more inclusive linguistic
conventions of written texts; and the third is the encompassing constraints of the
rhetorical problem itself. Writing is like trying to work within government regulations
from various agencies: Whatever the writer chooses to say must, in principle, eventually
conform to all of the constraints imposed from all of these areas. Let us look at each of
these kinds of constraints in more detail.
3.2. Knowledge
Generally speaking. Knowledge is a resource, not a constraint. However, it becomes
a constraint on the process when it in not in an acceptable form. In general, expository
writing calls for relatively orpnized. conceptually integrated knowledge. When
confronts n:; r. new or a complex issue, writers must often move from a rich array of
unorganized, perhaps even contradictory percentions, memories, and propositions to an
integrated notion of just what it is they think about the topic. Some writers obviously go
much further down this road than others, but much of the work of writing can be the task
of transforming incoherent thought and loosely related pockets of information into a
highly conceptualized and precisely related knowledge network.
In the following protocol, we se a subject responding to the demand for sufficiently
integrated knowledge. She has probably never had to talk, much less write, about her
subject before. so her writing process is strongly constra.ned by the need to formulate
just what it is sne thinks or knows. We see her retrie.fing information from memory,
drawing inferences, and relating her various ideas We have deleted portions of the
protocol that are irelevant to this discussion, they will be shown later. There are a
number of important tnings to notice here. .1f w try to .diagram the writer's developing
knowledge structure as a map, we find that the . ipography keeps changing. The writer
doesn't start with a well-formed thesis that she can just develop. Instead. she must
27
24
juggle her ideas around trying to decide just how they are related. "Grades" is an
interesting floater. Notice hnw it moves about on her knowledge map.
The arrows in Figure 3.2 indicate 3 general causal relationship between two ideas.
If that relationship become^ further specified, the line then receives a label as in episode
3. Initially both Grades and Prec_suia are linked independently to Motivation (lines 1-4 in
the protocol). Then Grades become identified with Pressure and subordinated to a new
notion, Personal Satisfaction. In episode 3, line 9 in the protocol, Personal Satisfaction is
reasserted as a cause of Motivation and the relationship between the two is further
defined with the label major, In episodes 4 and 5, lines 15-28 in the protocol, the writer
sets up a number of trial relationships in which Grades are still a subordinate element.
When, however, we skip to the final draft, we find a knowledge map in which Grades and
Personal Satisfaction have come to stand as independent parallel causes and each
relationship has been further specified by the labels major, and initial.
Insert Figures 3 1 and 3.2 about here
Retrieving knowledge and creating an adequate conceptual structure of "what you
think" can be a demanding task. Sir Phillip Sidney's poetic advice to A trophel, "Look into
they heart and write: is often a useful heuristic, but it doesn't guarantee that you will
find a ready-made conceptual structure there.
3 3 Written Speech
If we refer to the Wendy protocol at line 11 in Figure 3.3, we can see her trying to
, accommodate a second, even more demanding constraint. In addition to clarifying what
she thinks, she is now trying to express that knowledge map within the linguistic and
discourse conventions of written prose. Notice too how quickly she has jumped to the
added task of producing text. nine lines of analysis and she is ready to set it in type.
28
25
Insert Figure 3.3 about here
There are many ways in which language, which enables us to express complex
thought, also constrains our attempt to do it. For the inexperienced or remedial writer,
the rules of grammar and conventions of usage and syntax may make an enormous
demand on time and attention. But even the more experienced writer must encounter the
inevitable truculence of language itself, which seems to resist our attempts to form a set
of continuous sentences with forward and backward reference. A sentence that is
grammatically acceptable may twist the meaning, repeat a word too soon, or have terrible
rhythm. In generating a given sentence. the writer needs to meet all of these constraints
more or less at once
The following example illustrates the difference bet% een knowing something and
trying to turn that knowledge into a piece of writing. Wendy has established a knowledge
map in which Motivation and Grades are related in three distinct ways. She is now trying
to turn that set of thoughts into an acceptable sentence. Where we enter the protocol,
she is working on the sentences that will become sentences 2, 4, and 6 in the final text.
The excerpts shown in Figure 3.4. from Wendy's final essay and from the protocol.
illustrate two interesting points:
1. Complex thoughts don't automatically flower into appropriately parallelcomplex sentences. Although Moliere's Bourgeois Gentleman was surprised todiscover that he had been speaking *prose" all his life, doing so is no meantask The success that sentence-combining exercises claim for improvingoverall writing skill (O'Hare, 1973) is probably due to their ability to reduce theeffect of this linguistic constraint. By making sentence production processessomewhat more automatic, the writer has time to concentrate on otherimportant constraints.
2 In addition to producing a verbal rendition of thought, our writer must alsowork within the conventions of written speech, particularly those conventionsthat distinguish oral speech from writing and make writing a specialized formof discourse. Even from this brief protocol passage, we can infer that the
26
writer probably has a set of rules or adages about paper writing that say:
* Be specific.
* Repeat ideas for emphasis.
* Refer back for coherence.
* Don't repeat words/phrases in close proximity.
* Use "correct" (7) wording.
Insert Figure 3.4 about here
4. Cognition of Discovery
Even though the teacher gives several students the same assignment, The writers
themselves create the problem they solve. Because people only solve the problems they
give themselves, the act of representing the problem has a dramatic impact on
performance. People simply rewrite an assignment or a situation to make it
commensurate with their own skills, habits, or fears (Britton et, al., 1978). Although
writing texts generally ignore this part of the writing process, (Larson, 1978) our work
suggests that it may be one of the most critical steps the average writer takes.
The first part of this section, then, will describe our method for studying the
cognitive .rocess by which people represent the rhetorical problem. Then we will
present a model of the rehetorical problem itself. that is, a description of the major
elements writers could consider in building such an image Finally, we will use this
model of the possible as a basis for comparing what good and poor writers actually do.
30
27
4.1. Studying Cognitive Processes
The research question we posed for ourselves was this: if discovery is an act of
making meaning, not finding it, in response to a self-defined problem or goal, how does
this problem get defined? Specifically, we wanted to answer three questions:
1. What aspects of a rhetorical problem do people actively represent tothemselves? For example, do writers actually spend much time analyzing theiraudience, and if so, how do they do it?
2. If writers do spend time developing a full representation of their problem,does it help them generate new ideas?
3. And finally, are there any significant differences in the way good and poorwriters go about this task?
In order to describe the problem definition process itself, we collected thinking-aloud
protocols from both expert and novice writers A protocol is a detailed record of a
subject's behavior. Our protocols include a transcript of a tape recording rnsde by writers
instructed to verbalize their thinking process as they write, as well as all written material
the writer produced A typical protocol from a one-hour session will include four to five
pages of notes and writing and 15 pages of typed transcript. The novice writers were
college students who had gone to the Communication Skills Center for general writing
problems such as coherence and organization. The expert writers were teachers of
writing and meiotic who had received year-long NEH fellowships to study writing. Each
writer was given the followihg problem "write about your job for the readers of
Se.£nteeri magazine, 13-14 year-old girls," and was asked to compose out loud into a
tape recorder as he or she worked. They were told to verbalize everything that went
through their minds, including stray thoughts and crazy ideas, but not to try to analyze
their thought process, just to express it.
28
4.2. A Model of the Rhetorical Problem
From these protocols, we pulled together a composite picture or model of the
rhetorical problem itself. This composite is shown in Figure 4.1, with examples drawn
from our writers' protocols. It is based on what the group of writers did and shows the
basic elements of a writing problem which a given writer could, active;u consider in the
process of composing, if he or she chose to. For example, the writer in the following
excerpt is actively creating an image of himself or his persona, an image of what effect
he might have on his reader, and an initial representation of a meaning or idea he might
choose to develop, as the words in brackets indicate.
Ah, in fact, that might be a useful thing to focus on, how a professor differsfrom...how a teacher differs from a professor, (meaning), and I see myself as ateacher, (persona), that might help them, my audience, to reconsider their notionof what an English teacher does. (effect on audience)
Insert Figure 4.1 about here
Taken as a whole, the rhetorical problem breaks into two major units. The first is
the rhetorical situation This situation, which is the writer's given, includes the audience
and assignment The second unit is the set of goals the writer himself creates. The four
dominant kinds of goals we observed involved affecting the reader, creating a persona, or
voice, building a meaning, and producing a formal text.
4.3. Oifferencees Among Writers
This six-part model of the rhetorical problem attempts to describe the major kinds
of givens and goals writers could represent to themselves as they compose As a model
for comparison it allowed us to see patterns in what Our good and poor writers actually
did The differences, which were striking, were these'
1. Good writers respond to all aspects of the rhetorical problem. As they compose
29
they build a unique representation not only of their audience and assignment, but also of
their goals involving the audience, their own persona, and the text. By contrast, the
problem representations of the poor writers were concerned primarily with the features
and conventions of a written text, such as number of pages or magazine format. For
example, Figure 4.2 shows a vivid contrast between an expert and novice when we
compare the way two writers represented their rhetorical problem in the first 60 lines of
a protocol. The numbers are based on categorizing pharses and sentences within the
protocol.
Insert Figure 4.2 about here
As you can see. the expert made reference to his audience or assignment 18 times
in the first seven to eight minuses of composing. whereas the novice considered the
rhetorical situation less than half that often. The most striking difference of course, is in
their tendency to represent or create goals for dealing with the audience. Finally, the
column marked "Total" shows our expert writer simply spending more time than the
novice in thinking about and commenting on the rhetorical problem, as opposed to
spending that time generating text
2. In building their problem representation. good writers create a particularly rich
network of goals for affecting their reader Furthermore, these goals, based on affecting
%
a reader, also he!ped the writer generate new ideas In an eerier study we discovered
that our experienced writers fa different group this tape) generated up to 60 per cent of
their new ideas in response to the larger rhetorical problem (that is. in response to the
assignment, their audience, or their own goals) Only 30 per cent were in response to the
topic alone For example, a writer would say "Ill want an introduction that pulls you in,"
instead of merely reciting facts about the topic, such as As an engineer the first thing to
do is .." In the poor writers the results were almost reversed. 70 per cent of their new
30
ideas ?ivere statements about the topic alone without concern for the larger rhetorical
probltm (Flower and Hayes, 1979). All of this suggests that setting up goals to affect a
reader is not only a reasonable act, but a powerful strategy for generating new ideas and
exploring even a topic as personal as "my job."
As you might easily predict, plans for affecting a reader also give the final paper a
more effective rhetorical focus. For example, one of the novice writers, whose only goals
for affecting the audience were to "explain (his) job simply so it would appeal to a broad
range of intellect," ended up writing a detailed technical analysis of steam turbulence in
an electrical generator. The topic was of considerable importance to him as a future
research -ngineer, but hardly well focused for the readers of Seventeen.
3. Good writers represent the problem not only in more breadth, but in depth. As
they write, they continue to develop their image of the reader, the situation, and their
own goals with increasing detail and specificies. We saw this in the writer who came
back to revise and elaborate her image of her fashion-consuming reader. By contrast,
poor writers often remain throughout the entire composing period with the flat,
undeveloped, conventional representation of the problem with which they started
The main conclusion of our study is this: good writers are simply solving a
different problem than poor writers Given the fluency we can expect from native
speakers, this raises an important question Would the performance of poor writers
change if they too had a richer sense of what they were trying to do as they wrote, or if
the had more of the goals for affecting the reader which were so stimulating to the
good writers? People only solve the problems they represent to themselves. Our guess
is that the poor writers we studied possess verbal and rhetorical skills which they fail to
use because of their underdeveloped image of their rhetorical problem. Because they
have narrowed a rhetorical act to a paper-writing problem, their representation of the
problem doesn't call on abilities they may well have.
34
31
5. A Taxonomy of Writing Plans
We know that writers generate an enormous number and variety of plans as they
compose; the problem is how to categorize these plans in a useful way. Our hypothesis
is that writers draw on three major kinds of plans which are hierarchically related to one
another.
Si. Plans To Do
To begin with, writers generate plans for dealing with their rhetorical problem.
These rhetorical plans are called plans To Do something in or by language. These are
essentially plans for performing a speech act--for responding in some way to that
rhetorical problem. which includes the writer, the reader, and a purpose. A plan To Do
something in writing might be as unique and specific as "Write a note for the icebox door
to keep the family our of the plums Use a stern parental voice that begins with firm
reasonableness and ends with a veiled threat' At the Other extreme a rhetorical plan
could be as conventional and limited as "write another essay for Freshman Compositions
class." As you might expect when writers fail to plan or depend on limited, stereotypic
plans, they are likely to spend very little time actively considering audience or purpose
when they write They are more likely to produce "Writer-Based prose," which takes on
the structure of the ihriter's own thought process and the style of an interior monologue
(Flower, 1979).
A rhetoric& plan To Do something can not only improve the quality of a paper. it
can also make it easier to write When people treat writing as a speech act, they are
more likely to draw on many of the well-learned strategies adults use everyday for
arguing, explaining, or describing, but which many seem to ignore when they are writing
for a class A rhetorical plan offers the writer a pole star for the choppy sea of trying to
compose
35
32
5.2, Plans To Say
In order to carry out a plan To Do something, writers often generate two kinds of
subplans. The first of these is the familiar and rudimentary plan that all schoolchildren
have had drilled into them in the form of outlining: a plan for what you want To Say A
plan To Say something is essentially a content plan--a simplified or abstract version of
the information you want to convey. It can take a variety of forms, ranging from
scribbled notes and sketches on an envelope to an impressive sentence outline complete
with Roman numerals and two subpoints under every point. A plan To Say is essentially
a scale model of the final product. Perhaps that is why it has been so widely and
rigorously taught, often to the exclusion of any other kind of planning.
5.3. Composing Plans
There is, howe4er, another kind of planning writers do that is based not on the
product of writing, but on the process. This third kind of plan we call a Composing Plan.
Some Composing Plans help people generate knowledge. In classical rhetoric,
such formali2ed plans go under the name of invention. One can choose from highly
swstematic and analytical plans, such as the particle. wave, field analysis of tagmemics, to
Aristotle's topics or Gordon's synthetics. Or one could choose from more enigmatic and
inspirational plans, such as Sheridan Baker's (1969) advice on 'picking an argument' or
the meditation techniques used in Pre-Writing. on down to the time honored methods of
poetic inspiration look into your heart and write." If you wish your students to have
mo:e self conscious control over the process of generating ideas, there are many ways
to teach it.
The category of Composing Plans also includes a large set of ad hoc plans people
use to guide themselves through the process of writing For example, when the w. r in
the Wendy protocol ran into trouble, she told herself to "write a bunch of ideas down and
connect them later.' Some of our subjects appear to be at the mercy of inspiration as
33
they compose, or slaves to their own growing text. Others are able not only to monitor
their composing process, but to choose alternative ways to proceed. At the base of our
work with heuristic strategies for writers (Flower and Hayes, 1977) is an attempt to learn
more about these unexplored alternative strategies within the art of composing itself.
Let us close with an excerpt from a protocol that illustrates a writer working under
a top-level plan To Do something, which in turn creates a nested set of goals and
subgoals. As the protocol develops, we see how the writer's forward progress is the
result of a recursive, nonlinear process guided by a variety of plans. As an illuminating
contrast to this Subject, we studied another writer whom we shall call "Free write." As he
composed, Freewrite's top-level plan appeared to be 'Write whatever comes to mind"
His guiding plan was essentially a plan To Say, with only a rudimentary set of composing
rules tacked on (e.g, use correct grammar, use correct spelling if you know it, and
paragraph occasionally) His protocol showed almost no discernible attention given to
audience or purpose, and the final product, as you might guess, read rather like a
audience or purpose, and the final product, as you might guess, read rather like a
transcript of free association, even though the writer considered it quite adequate.
Insert Figure 5.2 about here
We return then to the writer working under a top-level plan To Do This schematic
version of a protocol covers the beginning of the composing session The plans To Do
and To Compose are generally comments the ,vriter makes to himse'f, whereas the plans
To Say are frequently notes jotted on paper Notice how the first three moves essentially
define the rhetorical problem.
By move 4. the writer has sketched out the rhetorical problem (his purpose,audience, and his own role) and set up a composing plan (just jot things down).When he begins to explore his knowledge at move 4, it is under thesimultaneous control of these two plans
37
34
Move 5, a decision to keep on generating ideas, is a reaffirmation anddevelopment of the initial composing plan in Move 2.
By move 7, the information the writer has generated leads him to form a newplan that is both a Composing plan for the final paper and a plan To Dosomething--to make a point for the reader.
Move 9 is probably the most illuminating point of the protocol because thewriter encounters a mismatch between his Knowledge (things he could sayabout Memory Is) and his goal vis-a-vis the reader. His action demonstratesthe distinction between Knowledge and Goals in writing. His high-level plan ToDo, based on his purpose and reader, lets him consider two subplans (make thesubject itself important or focus on its underlying principle) and in turn twopockets of knowledge. In the process of working by plan our writer considerstwo radically different things he could say. Clearly his writing process is notsimply the straightforward act of expressing what he knows. instead it is ahierarchically organized, recursive process in which knowledge and text aregenerated under the direction of both the rhetorical plan To Do something and aComposing plan for how to do it in writing.
This fragment of protocol was the beginning of the Subject's writing session. At
the end of the session. 40 minutes later, there was an unexpected code The writer
discovered that his initial objective of "justifying Memory r hl been entirely forgotten in,..
the course of composing a different line of argument. He now sees that Memory I (and
the ideas generated in our excerpt) could be an example in this larger argument. In the
following brief section, he sets up a new plan (which is both a rhetorical and a
composing plan) and begins to compose text.
This excerpt illustrates what is probably one of the critical differences we have seen
between the processes of good and weak writers. Weak writers in this situation *Mould
probably cont,nue to crunch out text under the direction of a plaii To Say what they,knew
or a plan To Compose their information into "acceptable" text. Good writers, by contrast,
not only make initial high-level plans To Do something, but continue to return to and
develop those plans as they write
Insert Figure 5 3 about here
35
6. The Pregnant Pause
An important aspect of most writers composing plans is the heuristic of
fractionation (Hayes. 1981). Fractionation is the process of breaking a problem into parts
and solving it by solving its component parts. the power of this heuristic for reducing the
impact of problem constraints and memory limitations is widely recognized. The writer's
use of fractionation to solve writing problems is revealed in the structure of the thinking
aloud protocols. Typically, writing protocols are divided into easily perceived segments or
"composing episodes" which are with few exceptions devoted to the statement and
solution of a relatively well defined part of the total writing problem.
Composing episodes are units in thi process of the writer, rather than his or her
written rolt,_.ict. We initially noticed that writers appeared to work in units of
concentration or periods of sustained focus, and, more importantly, found trial'
boundaries between these composing episodes could be agreed upon by independent
readers. In the protocols of three subjects analyzed in detail (the tape of one expert
writer was no longer available) these episodes ranged in length from 1 to 33 tines of
typescript, lasting from 7 seconds to 12 minutes, with an average length of 1 minute 45
seconds and an average of 10 clauses per episode (see Table 6 1).
Insert Table 6 1 about here
in our analysis we will occEsionally separate episodes into "major episode . which
are clearly autonomous episodes with strong boundaries, and "minor episod which
have weaker boundaries or stronger connections to adjacent episodes. ets of these
minor episodes typically cluster together to form a functional unit--they work as sub-
episodes within the more clearly bounded unit of a major episode The following example
will clarify this distinction. It comes from the very beginning of an expert protocol and
shows two brief major episodes (74 and 47 seconds) and the beginning of one longer
Ci
36
major episode (13 minute) composed of three minor episodes. The bo ridaries between
major episodes are indicated by double slash marks; those between minor episodes by e
single mark.
There r a number of features worth mention here. First, if this writer's
performance were merely observed, it would appear to be a long 160 seconds of pausing,
broken only by the act of shutting the door and ended by the transcription of the first
sentence. The protocol, however, reveals a substantial and complex body of planning.
Even during verbalizing, pauses still occur and those of 2 seconds or more are noted with
superscript numbers. The coding of the protocol reflects our model (Hayes and Flower,
1980, Flower and Hayes, in press. a) and distinguishes between the processes of Planning,
Translating (producing written text). and Reviewing Here text produced by the writer is
underlined once, reading is underscored twice.
Note that the first episode ends with a metacomment--a familiar enough ploy for
diverting attention from the task at hand. Episode two begins with a renewed attack on
the assignment, which told subjects to work as if they were free-lance writers. The third
episode breaks into smaller internal units or minor episodes. Like many of the episodes
focused on the act of Translating or producing prose, it is relatively long and broken into
minor episodes by brief evaluative comments and attention to side issues; yet the thread
of composition is not lost. Such episodes seem directed by an overall plan that can
sustain changes'rn topic and can cross paragraph boundaries. In this case, notice how
adroitly the fragment of test produced in Episode 3 responds to the audience analysis.
implicit goals. and plans which preceded it. As we will try to Show in the rest of this
paper. these episodes are goal-directed planning units in which writers work.
37
6.1. Evidence for an Episode Structure
Our analysis of the 'ontent of "pregnant pauses" will rest on two assertions we
have attempted to verify:
1. That the "composing episodes" are real; that is, that they represent meaningfuland verifiable units of ceocentration in which writers normally work.
2. That by looking at the bour(daries which occur between episodes we are infact looking at many of the longer, "pregnant" pauses noted by overvationalresearch.
Although these episode patterns had intuitive validity to rea_lers of the protocr4s,
we looked or converging evidence to support their reality and the reliability of our
boundary judgments. One content free indication of a boundary comes in the form of
signal words such as "all right," let's see." and "okay." In three of our four writers these
signal words clustered significantly (p< 001) around episode boundaries (i.e.. appearing in
either the immediately preceeding or succeeding clauses). The fourth writer simply didn't
use signal words. Such expressions seem to indicate a sense of opening or closura as in
"okay, now.." or "that's okay." Or they worked as filler in the writer's t ansition from one
episode to the next. (Although many boundaries are sharply defined points. some are one
to three clause transitional units containing false closures, false starts. and
metaccmments.)
Moct of the evidence for episodes and their boundaries, however, depends on the
m:ernal logic of the protocol itself As Table 6.2 shows. when judges study the protocol
carefully and know it well. they achie,e high reliability in judging boundaries These
knowledgeable" judges were merely instructed to look for units of coi.:entration in the
writer's process and to mark a boundary when they saw the writer shifting it.icus.,
changing a train of thought, or setting up a new plan. These judgements did show a
threshold effect (as verified by the Gutman scaling technique)--some judges simply had
broader criteria for selecting boundaries and noted more of therm However, even with
38
these differLnces, out of a total of 248 boundaries selected by our four knowledgeable
judges. two or more judges agreed on 70% of these boundaries. A random selection
predicted by a multinornial probability test would have yielded only a 20% agreement.
Insert Table 6.2 about here
For comparison, the protocols were then judged by four more readers whom we
called "intuitive judges" because they had not studied the protocols and were given no
instructions beyond "Use your intiution to mark what you see as meaningful episodes in
the process of the writer's thought." As a control they were given a set of markers
slightly greater than the number of major and minor episodes noted by the
knowledgeable fudges As one might expect, the intuitive judges created many more
idiosyncratic boundaries (i.e., those chosen by only . ie judge). Neverthel&.s, of the 290
boundaries they marked, two or more judges agreed on 50% of the boundaries. They~I.
showed even stranger agreement an the "official" boundaries (i.e, those selected by two
or mi., 3 knowledgeable judges). Two or more of the intuitive judges seinted over 90%
of these boundaries (whereas, a probability test would predict only 13%) and three or
more fudges agreed on 73% (compared ) the even smaller probability prediction of only
.5% agreement) Finally, we asked a group of twenty-two writing researchers attending a
seminar on protocol analysis to make intuitive judgments on yet another protocol and
found that eight readers or more agreed on 70% of all the boundaries chosen by the
group (here probability would predict such agreement on only 00000118% of the
boundaries). We think these results are remarkably strong.
The goal of this initial analysis was not to create a well-specifir.d definition of
episodes, but simply to gather prima face evidence that they do indeed exist as complex,
yet strongly visible units within the composing process. We can sum up he findings in
this way You can't expect every reader to agree on all the boundaries; yet major
39
episode boundaries have a high intuitive discernibly. The selection of minor or sub-
episode boundaries will be more Idiosyncratic since readers respond to a wide range of
events such as precess shifts between planning, translating, and editing; shifts in topic;
and the intrusion of metacomments. However, knowledgeable readers--that is, people
who carefully study the content and logic of the entire protocolwill come to high
agreement in choosing episodes. The importance of "knowing" the protocol also reveals
a key fact about episodes. Episodes are not like paragraphs of a text, organized around a
central topic which a casual reader can easily follow. Instead, episodes seem to be
organized around C#031S, so that one episode could include various topics and various
processes from planning to editing- -all tied together by their relevance to the writer's
current plan or goat Readers who know the protocol well are more aware of this overall
structure
7. Evaluating How Writers Generate Ideas
An important part of the planning process as it is described in our model is idea
generation. In this section, we present data which helps us to describe idea generation
more fully.
This study started with the hypothesis that an important difference between good
aria poor writers lies, not simply in their ability to express ideas in written speech, but in
the very strategies they use to generate those ideas in the first place We had observed
that poor writers. in their attempt to find a focus or thesis for a paper, often seemed tied
to the topic. while more expermoced writers appeared to be responding to a larger
rhetorical problem--a problem which included the reader and their own goals (Flower and
Hayes, 1980).
If this hypothesis were true it would mean a number of things. First, if poor writers
are obsessively focused on the topic to the exclusion of the larger rhetorical problem, it
could help explain why they often are more likely to violate conventions of the
WVaarlexImQW11111111111111111101111
40
appropriate when they write, but not when they speak. In face-to-face conversation oniy
ancient mariners are likely to ignore the rhetorical situation. Secondly, it the rhetorical
situation itself helps stimulate plans and ideas. it could explain why so many poor writers,
including the ones in our experiment, often seem to "run out of ideas." Finally, if a
significant difference between good and poor writers is the strategy they use to generate
ideas, this would suggest that evaluating or editing the final product gta product is
unlikely to produce dramatic change. A more effective teaching technique would focus
on the writing process itself.
To test this hypothesis we collected, on tape, verbal protocols of nine writers
composing aloud. Four were people on the university staff who liked to write, who had
done writing, and who were considered by their peers to be "good" writers In contrast
to this group of "good" or experienced writers, we studied a group of "poor" writers who
had come or been sent to the Communications Skills Center for general problems with
organization and coherence. Two of these "poor" writers were graduate students, two
undergraduate, none had difficulty with basic grammar or sentence structure
Each subject wrote on an assignment about which they would have topic
info,r,ation, but which created significant audience constraints They were told to work
foe approximately an hour and to verbalize everything that went through their minds as
they wrote We analyzed these transcripts in two steps first, by isolating each new idea
that was generated during the session (see Table 7.1). A new idea was defined as any
Complete grarnmeical unit, including complex statements with dependent clauses
However, if such a complex statement was generated in two attempts separated by long
pauses or intervening material, it was coded as two ideas. Our goal was to code as one
unit those ideas which were being retriced from memory as a unit, and to code as new
all new attempts to expand or develop an idea Changes which affected merely the
wording or sentence structure were budged on the basis of our model of the writing
1
41
process to belong to the process of "translating," not "generating," and were not counted
(Hayes and Flower, 1980).
Insert Table 7.1 about here
Our second step was to discover where each of these new ideas came from.
Therefore we categorized each new idea in one of three ways, as a response to either:
1. the larger rhetorical problem, including the topic,
2. the topic alore,
3. a current element in memory.
The basic purpose of this categorization was simple. Within a given body of ideas,
we wanted to see what per cent of those ideas were generated as a response merely to
the topic alone (or to a current element in memory) or alternatively, as a response to not
only the topic, but to the larger rhetorical situation as well.
We used the following taxonomy to decide into which category a new idea should
be placed:
1. An idea was categorized as a response to the rhetorical problem if it indicatedone of the following: a concern with the writer's Purpose or Goal; anindication of the writer's sense of Audience; or a concern with the writer'ssense of Self or Persona. A writers concern with purpose or Goals took twomajor forms either as a statement of purpose (e g , "I need something herethat pulls you in") or as a recognition of some of the formal features ofwritten text Statements such as, "I'll use this as an introduction," indicatedthat the writer was see.ng her ideas in the larger contaAt of writing a paper orarticle Ideas which responded to the audience weie sometimes direct ("MakethiS friendlier; its for a young audience") or implicit (e g.., "I'll list the names ofthe most fascinating drinks'). Writers generated ideas in response to theirprojected Self or Persona with comments such as. "I'll appear like an idiot."Finally, some statements combined a number of these elements, such as "Thismay not be the best term for ten-year-olds, but it maintains the rhythm."Any idea which showed some response to the larger rhetorical problem. then,was placed in the first category
2 The second category was reserved for new ideas which were judged as
42
simply information generated in response to the topic alone, such as "Awaitress has a number of duties, first...." Often these ideas appeared to bethe result of a straight memory search of what the writer remembered aboutthe topic.
3. The final category was necessary to account for new ideas which appeared to, be connected by some association to a recent thought or current item in
memory, but which were not relevant to the rhetorical problem or to thetopic. Both good and poor writers appear to go off on these short trains ofassociation; the difference is in how frequently they do it.
As we expected, it is difficult to get complete agreement from judges on the
absolute number of new ideas, since different judges are likely to have 'ifierent
thresholds for distinguishing new ideas from mere rewordings of old ones. Therefore, as
a check on our method we conducted the following test using a set of 73 ideas which
were judged as new ideas by all four judges in the experiment. The critical judgments in
this study are the categorizations: is a gil...1 idea a response to the rhetorical problem or
merely to the topic or a current element? And can judges agree on making those
categorizations? In a universe of 73 ideas which all four judges had selected .a.s new
ideas. there was complete categorization agreement on 52 ideas or 71%. Three judges
agreed on 69 responses or 94.5°/e. There was a two or three-way split on only four ideas
This percentage of agreement confirmed our sense that such categorization is reliable
7.1. Results
The results of this analysis were striking, especially since our subjects did not
represent the extremes of either good or poor writers and there was not attempt to
account for or control individual differences. Nevertheless, as Table 7.1 shows, the poor
writers as a group generated on 28% of their new ideas in response to the rhetorical
problem, the other 72% were in response to the topic and/or a current element in
memory. For the good writers this 30/70 distribution was nearly reversed Good writers
generated 60% of their new ideas in response to the rhetorical problem in some way;
only 400/ci of their ideas were a response to the topic or current element alone.
43
As you can see in Table 7.1, there is a significant spread among the poor writers,
but as a group they remained distinct from the good writers. On the basis of a Pitman
randomization test, these results were significant at the .01 level; that is the probability
that this difference etween the groups would have arisen by chance is less than 1 in 100
(Siegel, 1956).
8. Formulating Sentences in Writing
In the previous several sections, we have examined the planning process, the first
major process in our writing model in considerable detail. In this section, we turn to the
second major process, Translation
How do people actually write sentences? When asked ourselves this question, we
found that there was a great deal we didn't know We knew, of course, that writers' plans
are very important, but we weren't sure of the extent to which the writers' plans
determined the details of the sentences written nor of the extent to which the experience
of composing sentences modified the writers' plans We didn't know if sentences are
composed as a whole or if they are assembled from separately composed parts Further,
we didn't know if there were differences in the processes experts and non-experts use to
compose sentences
This paper explores all three of these questions. In the first section, we will discuss
the relatpon of planning and sentence composition. In the second section, we will present
data on the processes by which writers compose sentences We will propose a model to
:iccount for the data In the third section, we will describe differences in the way experts
and non-experts compose sentences
In conducting this research. we were guided by our model of written composition
(Hayes and Flower, 1880) In this model (see Figure 2.1 Section 2), we proposed three
major writing processes. PLANNING, TRANSLATING, and REVIEWING The function of the
PLANNING process is to set goals and to establish a writing plan whith will guide tte
47
44
production of text. The TRANSLATING process acts under the guidance of the writing
plan to produce written text. The function of the REVIEWING process is to improve the
quality of the text.
Below, we will present observations of people writing formal sentences, and use
these observations to develop a detailed model of the TRANSLATION process and its
relation to PLANNING.
8.1. The Relation Between Planning and Sentence Construction
We observed writers as they wrote essays on topics such as "My Job- and
"Abortion: Pro and Con- for teenage audiences. The writers were asked to "think aloud"
while writing. The resulting protocols were tape recorded and transcribed Our subjects
were six expert writers chosen for their professional involvement in writing and six
subjects who were competent adult writers, but not professionals.
For some writers, planning prior to writing is very sketchy, apparently consisting of
little more than the choice of a general topic or perhaps a decision to write in simple
language. Others plan more extensively--developing lists of subtopics to be discussed in
a particular order--before any sentences are written.
Plans influence the way in which sentences are written. The order in which topics
are discussed in the final essay is typically closely related to the order in which they are
listed in the notes the writer made during planning.
The left nand columr )f Figure 8.1 shows the plan for oryanizing the essay by one
of the expert writers as revealed by his protocol and by his notes written before he
began to construct formai sentences. The right hand column shows the sequence of
topics and subtopics in this completed essay.
Insert Figure 81 about here
45
For this writer, the relation between plan and essay was very close.
One output of the PLANNING process is typically a sequence of brief written notes
of two kinds:
1. topic designations, e.g., -religious reasons," misery of being an unwantedchild: and
2. instructions to the writer, e.g., "introduction," "snapper line."
These two kinds of notes seem to serve rather different functions Topic designations
serve to remind the writer to include information about a particular subject matter. Often
it appears that the writer has this information preorganized for presentation in long-term
memory. Instructions to the writer, on the other hand, remind the author to accomplish
some rhetorical goal such as providing an app:opriz.v.: beginning or end to the essay.
Often they are content free or direct the writer to organize information at some later
time For example, an instruction such as "summary" will typically lead the writer to
organize content for a summary only after the essay has been written. .....
When the writer begIns to generate formal sentences, these brief notes will be
greatly expandea For the writer whose plan and essay are described in Figure 8 1, each
plan element gave rise on the average to 2 5 sentences and about 55 words. In word
count, the topics in the outline were expanded ahOut eight-fold on the average when they
appeared in the finished essay Expansion varied from a low in which a 29 word note
was transformed into 39 ..fiords of text to much more extensive expansions in wnich, for
eAarn0e, "snapper line" became a 53 word conclusion and the single word "age" gave rise
to 58 words of text.
As we noted above, our writers usually did not make outlines as complete as this
water did As a result, the smount of expansion from outline to essay for most writers is
greater than we observe here.
For none of the writers we observed was the order of topics in the essay exactly
45
46
the same as the order of topics in the plan. Many changes in plans occurred while the
writer was composing sentences. The writer whose plans are outlined in Figure 8.1
decided after writing section 2a(2) of the essay, "economic reasons," that section 2b of
the plan, "current con-position," was "repetitious" and eliminated it. After writing section
2a(3), he decided that section 2c was unnecessary. Later he decided that topic 4b in the
plan was really two topics and wrote it as such in 'he essay.
Clearly, then, the writer's plans influence the construction of sentences.
Constructing sentences, however, can also influence plans. Just how a plan will work out
isn't always clear until the writer tries to execute it. When he tries to put the plans into
words, he may well discover weaknesses and redundancies which were not obvious and
perhaps could not be obvious earlier.
The process of writing sentences can lead to more than lust a change in the writing
plan. It can also provide the occasion for writers to change their understanding of the
topic. In the protocol segment shown in Figure 8.2, the writer is trying to compose a
sentence about writing difficulties.
Insert Figure 8.2 about here
At first the writer seems to feel that the important problem is impulsiveness. By
the time the sentence es Completed, hov,eer. he decided that the real problem is lack of
planning
Our maul point here is that even when writers do make compete plans, there is still
plenty of work to do to construct formal sentences. We know this because
1 The plan will almost certainly be modified during TRANSLATiON.
2 The plan w:11 be expanded ten fold or more to produce the sentence And
3 Some elements in the plan are instructions to the writer such as
,ntrOduction" or "snapper line" which require the writer to add content at the
5(1
47
time of writing sentences.
We can summarize the relation between the PLANNING and TRANSLATING process
as follows:
1. The order of topics in the writing plan is closely related to the order of topicsin the essay. The PLANNING process clearly exerts some control over theTRANSLATING process.
2. When writing formal sentences, the writer greatly expands the topicsdesignated quite briefly in the plan. The TRANSLATING process, then, takesthe plan as input and builds on it.
3. Writing sentences can lead the writer to modify the writing plan. TheTRANSLATION process, then, can exert some reciprocal control overPLANNING.
8.2. What Happens When Sentences Are Written?
The top part of Figure 8.3 is a protocol segment in which the subject was
composing and writing down the sentence shown at the bottom This segment shows all
of the important features of sentence generation that we have observed in our sample of
writers. First and most important, the subject constructs sentences by proposing and
evaluating sentence parts
Insert Figure 8.3 about here
Items 1, 4. 6, 9. 12, and 17 are proposed sentence parts. Items 10. 13. and 15 indicate
evaluations In addition, the protocol segment reveals three other processes:
interrogation, goal setting. and rereading Items 2 and 8 are interrogations. It is very
common for subjects while they are writing to ask themselves questions such as. "What
do I want to say?", "What do I mean?", "What did he do?", or simply, "What?" We assume
that these interrogations reflect memory search processes in which the writer is trying to
find information to be used in constructing the sentence.
Items 3 and 5 are instances of goal setting in which the writer specifies some
., 51
48
properties desired in the sentence without producing them.
Items 11, 14, and 16 are instances of reading.
The average size of proposed sentence parts for our 12 subjects was 9.29 words,
and the average number proposed was 2.78 parts per sentence. Of the words proposed
in,.the form of sentence parts, just over three-quarters (.76) were included in the final
sentence Thus the process of proposing sentence parts appears to be a fairly efficient
one
Of all the protocol segments which correspond to sentence construction:c
* 18% contain ;.ine or more interrogatives
* 66% involve rereading of previously currently being written, and
* 8% involve rereading of previously written sentences
Rereading of the current sentence, then, I., much more common during the TRANSLATING
process than rereading of sentences composed earlier.
Figure 8.4 shows a model of the TRANSLATION process which accounts for many of
the observed behaviors. The model assumes that in constructing sentences, the writer
will try to follow a sequence of plan elements formed earlier. If a plan element is
evaluated negatively, or if for some reason there is no plan element, then the writer will
initiate PLANNING. In both PLANNING and EVALUATING, the writer may and often does
make use of informaticn derived by reading the context of the sentence to be produced
Insert Figure 8 4 about here
Once a plan element is selected, the writer will attempt to express it by writing one
or more sentences This will be done by proposing and evaluating sentence parts If the
writer had difficulty in proposing a part, then she/he may reread the context of the
sentence ("to get a running start ") or may ask a clarifying question such as. "What did he
52
49
really do?" or. "What am I trying to say?"
If a proposed part is evaluated positively, it is added to the current sentence buffer.
We assume that parts are added to the sentence buffer from left to right and that at any
time the buffer contains the first or left hand part of a sentence. We believe that it rarely
or never contains a sentence part which is detached from the beginning of the sentence.
A sentence part may evaluated negatively because it fails to match the intended
plan. For example, the proposed part may state so. tething either more general or
something more specific than the writer intended. If the writer continues to have
difficulty in finding a part to add to the current sentence buffer, she/he may start over by
clearing the sentence buffer.
When a sentence is completed, the writer must decide if the current plan element
has been completed. If not. she/he will compose more sentences until the plan element
is completed If so, the write: will look for a new plan element.
$.3. Differences Between Experts and Non-experts
Generally, the protocols of the expert writers resemble those of the competent
writers Both groups construct sentences by proposing and evaluating sentence parts,
and both groups engage in interrogation, goal setting. and rereading of the current
sentence There are, however, two consistent differences between experts and novices7Experts write longer essays and experts propose longer sentence parts. Table 81 shows
ithe essay length and sentence part length for the experts and non-experts Both/
differences are significant beyond the .05 level by the Mann-Whitney test.
Insert Table 8.1 about here
An independent rater measured sentence part lengths for two of the subjectsS4
and S7--and obtained average part lengths of 12.23 words for S4 and 7 30 words for S7
5.3
50
The four experts who proposed long sentence parts (Subjects 1, 2, 3, and 4) wrote
essays which were ranked first, second, third and fourth by panel of judges who judged
all 12 essays. The experts who proposed short sentence parts (Subjects 5 and 6) wrote
essays which were ranked 9th and 10th in quality by these same judges. There is some
reason, then, to believe that the average length of the sentence parts which a writer
proposes when he is constructing sentences is related to the writer's skill in writing.
What mechanism could be responsible for such a relation?
Simon and Chase (1973), studying skin in chess, concluded that the advantage
which chess experts have over novices depends on a enormous amount of pattern
knowledge which they acquire during thousands of hours spent in analyzing chess games.
The fact that experts have more and larger patterns than those available to novices
allows the expert to think of chess games in larger units than novices can use. Perhaps
thousands of hours spent constructing sentences enables the expert writer to work in
larger units than those with less experience.
In contrast to the fluency shown by experts, the sentence constructing processes of
poor writers may be interrupted frequently by difficulties with low level processes. Figure
8 5 illustrates this sort of difficulty in a student who had been referred to a writing clinic.
Spelling, orthography, and even the simple matter of handling a pencil occupy so much of
the writer s attention that he has considerable difficulty keeping track of his sentence.
Cur competent and expert writers ramly ,.ave difficulties with low level processes
insert Figure 8 5 about here
In poor writers. then, fluency is strongly influenced by mastery (or lack of masteryl
of low level skills. in competent and expert writers, vie propose that fluency depends on
the acquisition of large quantities of sentence pattern knowledge.
54
51
9. Implications for Teaching
Our research on the composing process has a number of implications for teaching.
It sug,ests some important additions to what we teach: we need to teach students to
understand and analyze their own thinking process just as they now do their writing
products. And we need to reconsider how we teach, when the content of such teaching
is not content specific knowledge but Process skils. Teaching heuristics and thinkinn
strategies -- trying to affect performancecalls for new techniques. In summing up the
implications of our research for teaching, we would emphasize three important
observations supported by our work.
1. There are important differences in how expert and novice writers handle theprocess of writing.
2 Many of the heurstics or strate:lies experienced writers employ are emirentlyteachable.
3 One of the most promising areas for improving students' writing is in theneglected art of planning.
9.1. Implication 1
There are important differences in how expert and novice writers handle the
process of writing This difference was particularly evident in a study of how writers
definod their own rhetoncal prvblern- -what elements of the task they attend to and how
rich nd c( mph:. P "epresentation they built
1. Go .ers rAcpond to all aspects of the rt^"orical problem. As they compose
tr.dy build a ungive representation not only of their audience and assignment. but also of
their goals involving the audience. their own mane. and the text. By contrast. the
problem representations of the poor writers were concerned primari'y with the features
and conventions of a written text. such as number of pages or magazine format
2 In building their problem representation. good writers create a pe:rticularly rich
network of goals for affecting their reade Furthermore. the&.:s gals. based on affecting
55
52
a reader, also helped the writer ge. -'rate new ideas. In an earlier study we discovereo
that our experienced writers (a different group this time) generated up to 60 per cent of
their new ideas in response to the larger rhetorical problem. Only 30 per cent were in
response to the topic alone. For example, a writer would say, "I'll want an introduction
that pun you in," instead of merely reciting facts about the topic, such as As an
engineer the f'-et thing to do is . . . ." In the poor writers the results were almost
reversed 70 per cent of their new ideas were statements about the topic alone without
concern for the larger rhetorical problem. All of this suggests that setting up goals to
affect a reader is not only a reasonable act, but a powerful strategy for generating new
ideas and exploring even a topic as personal as "my job."
As you might easily predict, plans for affecting a reader also give the final paper a
more effective rhetorical focus For example, one of the novice writers, whose only goals
for affecting the audience were to "explain [his) job simply so it would appeal to a broad
range of intellect," ended up writing a detailed techr..cal analysis of steam turbulence in
an e!ectrical generator. The topic was of considerable importance to him as a future
research engineer, but hardly well focused for the readers of Seventeen
3 Good writers represent the problem not only in more breadth. but in depth As
they write. they continue to deelOp their image of the reader. the situation. and their
c...,,n goals with increasing detail and specificity We saw this in the writer who came
task to re :Ise and elab3rate her image of he.' fashion-consuming reader By contrast.
ccOr ...r.ters often remain throughout the entire compos.ng period with the flat.
unde,:etcped, COn4ent.onal representation of the problem with which they started
The main conclusion of our study is this good writers are simply solving a
different problem than poor waters Given the fluency we can expect from native
speat:ers. this raises an important iuestion Would the performance of poor writers
chan:e if they too had a richer sense of vhat they were trying to do as they wrote. or if
56'
53
they had more of the goals for affecting the reader, which were so stimulating to the
good writers? People only solve the problems they represent to themselves. 014r guess
is that the poor writers we studied possess verbal anc. rhetorical skills which they fail to
use becir..:se of their underdeveloped image of their rhetorical propie-n. By narrowing a
rhetorical act to a paper-writing problem. their representation of the problem doesn't call
on abilities they may well have.
The second implication we see in our own t:tady is that the ability to explore a
rhetorical problem is eminently teanabie. Unlike a metaphoric "ciscovery." problem-
finding is not a totally mysterious or magical act. Writers discover what they want to do
by insistently, energetically exploring the entire piablem before them and building for
themselves a unique image of the problem they want to solve A part of creative thinking
is just plain thinking
Exploring a topic *lone isn't enough As Donald Murray put it "writers wait for
signals" which tell them it is time to write, whi 1 "give a sense of tlosure, a way of
handling a diffuse and overwhelming subject." Many of the "signals" Murray described,
such as having found a point of view, a voice, or a gene, parallel our description of the
goals and plans we saw good writers making If we can teach students to explore and
define their own problems, even within the constraints of an assignment, we can help
them to create inspiration instead of wait for it
9 2. Implication 2
The process of writing is not a simple, step-by-step process However, many of
the heuristics or strategies which aid experienced writers are eminently teachable Three
implications for teaching I- -' iristics stand out:
I Heuristics do not offer a step-by-step formula for how to write. They areavailable. and powerful, but optional techniques for solving problems along theway Although it makes sense, in general, to plan before you generate and togenerate ideas before juggling them for a reader, these processes can often
57
54
be collapsed together in a writer'3 thinking. Furthermore, as our subjectsshow, the entire process and construct may be reiteratedtime and again at all levef of the kocess, from the act of articulating a keyphrase to producing a sentence, pargraph, or entire paper. Problem solvingasks the writer to trade 14 his/haf set of rules for :low to Write (Gather,Outline, and Write), which never worked too well aeiway. for a set ofAlternative Ways to Reach Your Goal When You Write.
2. A second basic ?act about teaching heuristics is that people must experiencia new thinking technique to learn it. Brainstorming, for example, is a9acquired skill and may ga against the grain for writers geared to producirigusable prose on a first sitting Students will not blithely relinquish theirhabitual coinnosing techniques, no matter how inefficient, at the sight of anew idea. To make a new heuristic an available option it must be presentedas a classroom exri, .ance which ensures that the writer actually learns howto use and apply a new technique Even the inexperienced writer is never atabula rasa. he comes equipped with many well-engrained, if counter-productive habits It is one thing to teach students a new formula. another toactually change behavior But writing. like problem-solving thinking ingeneral, is a performance art. Unless we deal with writing as a form ofthinking, we hove simply taught the student the ropes of another classroomgenre-the composition paper
3 Finally, a problem-solving approach to writing works for many writers becauseit allows for the disorderly dynamics of serious thinking and encourages onanalytical and experimental attitude in the writer. Heuristics ask the studentto see writing as a communication problem they are setting out to solve withall the strategies they can muster. In practice, perhaps the most remarkableresult of using heuristics is that early in the course students develop aconviction that writing is an important skill they can in fact master.Obviously, such a conviction is not always one hundred per cent warranted.but in replacing the mystique of talent and the fear of failing with thepossibility of an attainable goal, problem solving helps writers draw retire Sullyon the abilities they do have.
9 3. Implication 3
Finally, in teaching strazegies for the writing process, one of the most promising
areas is in teaching the often neglected are of planning
In a recent talk, Richard L Larson took English teachers to task for the way we do
and don t teach write! s to plan. According to his informal survey of cuisent textbooks,
our instruction in planning is limited to teaching a few old war horses and is focused
Quite decidedly on the written product, not the writing process. If students followed only
53
55
our teaching (and apparently it's a good thing they don't), the only kind of planning they"
would do would be limited to 1) making and outline, 2) choosing a method of
development and 3) deciding on transitions. In reali f, as Professor Larson points out,
writers also need to plan what they want to do in a piece of writing and where they want
-0..1 leave the reader.
Thzt statement seems an uncontestable piece of common sense. And yet, would
we agree with :t7 Is the common sense support of planning in conflict with the equally
reasonable assumption that writing is a process of discovery? Many of us would argue
that writers find where they are going on the way to getting there. And furthermore, thatJ
planning. especially the lockstep of an outline, can for.:.e a writer to leave the wandering
path of discovery in favor of mar:hing down a straight and narrow-minded path to t;ie
end of the theme According to t'le discovery method, planning may indeed help you get
there sooner, but the destination tou reach may not be worth the trip.
We could state the dilemma in this way. The act of producing a rhetorically
effective, purposeful piece of writing depends on highly goal-directed thinking, on making
plans On the other hand, the equally important act of making meaning where none
existed, of turning o,i experience into ideas, is a discovery procedure fostered by the
freedom to explore by-ways and follow unmarked paths that no plan could foresee
The practical problem for us as teachers is how to resolve this conflict, car. we give
students the power of planning without denying the experience of discovery, Textbooks,
insefar as they reflect teaching methods, often fan into one of these two camps.
emphasizing either hardline traditional methods of planning (outlines, methods of
. development. etc.) or discovery procedures such as pre-writing or free writing This
suggests that writing can be either an act of honest and creative self-exploration. or it
can be an act of planned , rhetorically effective problem-solving. But we seem to be
saying. it can't be both.
56
In contrast to this apparent dichotomy in teaching, research in composing process
suggests that good writers do both. In my own work with John R. Hayes studying the
thinking processes of writers, we see writers make plans to explore a topic, to discover
conflicts, to figure out what they really mean, and at the same time make plans to
produce written discourse and to deal with a reader. One of the important problems
writers face, but teachers teach, is how to map these various plans onto one another and
to coordinate exploration and communication to serve a common goal.
I suggest we often fail to teach this interaction for precisely the reasons Professor
Larson mentionedour definitions of planning are limited and limiting. Let me qualify
that statement. when we ourselves write, our working definition of planning might be
quite broad and flexible, it probably includes all those things that go on when one is
driving to work or standing in the shower But what we teacn under the name of plans
may still be outlines, methods of development, and transitions
The purpose of this paper is to briefly describe the kinds of planning writers do
which combine the power of goal-directed thinking with the richness of a discovery
process We will look at three kinds of plans. plans To Do Something by writing, plans
To Say Something in writing, and plans To Discover Something through the act of
writing itself But first let's take a brief look at the nature and power of plans
Contrary to the tradition of monumental sentence outlines glittering with Roman
numerals and two points beneath every sub-point, good plans are often only sketches in
the mind Plans help us write in three ways
1. Plans let people reduce large messy problems (such as be interesting") downto their essentials Architects do this when they create and revise sketchesinstead of experimenting in steel and concrete Writers do it when theychoose a focus, jot down notes on envelopes, draw pictures with arrows. orwrite outlines A plan, then, is a scaled down version of our solution to aproblem, a model which abstracts the essentials from a problem and allowsus to mentally manipulate those essentials first
2 This reveals a very important fact about plans Since a plan allows us to test
57'
out a solution in the way an artist uses a sketch, a good pLirl must bedetailed enough to test, but cheat) enough to throw away. That is why thoseelaborate early outlines often fait to produce good writing or help the writer.They are so expensive to create, they are less a plan than a shackle. Theylock writers into a premature solution before they have even entered theproblem. A good plan, then, is a sketch which sets up goals and alternativeswhich, in turn, keep the writer focused on the essentials of tne problem, notthe details of a particular solution.
3. Plans. at least some plans, have another characteristic. They give writers aset of steps or procedures for getting from where they are to where theywant to be. We say that good plans of this sort are operational; they help usact. One way to see if a plan is operational is to put it in the form of a goalstatement. Then see if it suggests how to proceed; if it offers built-in "how-to" cues for how to achieve the goal. Compare these two goals. one with"how-to" cues. one without: (1) I want to be rich and famous, versus (2)want to study probability, statistics, and writing so I can get rich quick at LasVegas and become famous writing a bestseller on how I did it. A writer mightmake a goal such as be persuasive" wore operational by saying, 1 want toargue forcefully both sides of this controversy to show the reader that I havepinpointed the crucial issues, but also to pave the way for my own ideas."Plans or goals without "how-to" cues are often highly abstract, for example, 1want to discuss team sports..impress my reader.. get an A in this course."Such plans may not offer the writer much help at all.
4. The third strength of plans is really a result of the first two. Because plansabstract a problem to its essentials and suggest ways to go about working onit, they help writers turn an overwhelming situation- -write that terrificpaper--into a manageable set of sub-problems By discovering andconcentrating on major sub-problems, such as the purpose of the paper,writers can handle each part better and reduce the anxiety of facing anunmanageable whole. As you might expect, good writers not only work onsuch sub-problems but have a variety of strategies for integrating the partsinto a whole For example, writers can delay consideration of a lower levelconcern such as spelling, grammar, or even organization until they haveworked out what they might want to say. But at the same time they continueto consolidate and reorganize what has gone before as their ideas develop
9.4. Teaching Planning Versus Doing It
Teaching has a lot in common with planning. We break a complex process down
into parts and teach people how to use the parts But inevitably. we leave the work of
integrating those parts into a whose up to the student The process of writing is too
complex to give anything like a recipe for It. But at the same time we must not confuse
61
58
the parts we teach with the process itself. This is particularly important with planning.
At the end of this paper I will offer three kinds of planning that we can teach as
independent planning exercises. In reality, of course, good writers use and integrate all
three kinds of plans which we as teachers must separate to teach. The question is, how
do good writers do that?
In trying to develop a model of such cognitive processes in writing, John R. Hayes
and I have used the method of protocol analysis to see what writers are actually doing as
they compose. In this research we ask writers to compose out loud, verbalizing
everything that goes through their minds as they are writing The transcript of this tape
recording, which is called a verbal protocol, along with the writers' notes and manuscript
provides an extraordinarily rich record of the thinking processes that underlie the act of
composing. When we look at the planning processes writers use. two things stand out
1 Plans do not emerge fully blown at the beginning of a writing session. Theyare often generated in response to the writers purpose, topic, or audiencePlans begin as sketches that get changed and fleshed out as the writerexplores the problem.
2 The planning process (that is, for the writers who do make plans, and somedon't) continues throughout the writing process. V'e may place planning atthe beginning of a textbook and encourage it at the beginning of writing, butit is not a formal exercise like outline making. It is a thinking activity, almosta frame of mind, that characterizes the entire writing process of good writers.So, even though we may teach the kinds of planning outlined below asindependent activities, we need to also make clear how they fit into the largeract of writing
With that in mind let me descrioe three planning techniques which try to bridge
research and teaching by translating what good writers do into teachable techniques that
help people write.
62
59
9.5. A Plan To Do
One of the most important but most untaught kinds of plans writers make are
rhetorical plans or plans To Do Something by writing. People write for a reason, and the
clearer they are about their goals the more likely they are to get there. In trying to
decide what they want To Do by writing, writers must define the rhetorical problem they
are facing: what do they think they are going to accomplish with whom, and how do
they think they are going to do it? Teaching students to do such rhetoric& planning, and
creating realistic assignments that require it, remind us that writing is a purposeful act
and not an exercise in style alone. Furthermore, rhetorical planning is an important way
good writers narrow down their search from all the possible things they could say about
a topic to the important things they want to say. Rhetorical planning simply makes it
easier to write well
Here is an example from a writer wno tried to map his plan for what he wanted To
Do by writing a letter to his Congressman. As you can see, it would have been easy to
simply write a list of facts about himself. Planning can escalate the problem The writer
must create and organize new concepts, not just 'print out" what he knows Such
planning sets new standards. and makes it possible to achieve them.
My purpose in vritIng is to convince the Congressman that I amthe best candidate for a legislative aid.
How do I cony Oa him that I'm the best applicant?
Show him that I am a personof purpose and determination
have set have set wayscareer goals to achieve them
college law school
Convince him that I possessthe skills that he would con-sider valuaole and hopefullybetter than other candidates'
analyticaland com-municationskills
govt. leader-experi- shipence roles
Another kind of Plan To Do writers often make is called an Impact Statement much
60
60
like the environmental impact statement a dam builder has to make, describing the effect
his dam will have on the land, wildlife, water and so on. An Impact Statement is focused
on the reader. For instance, the writer asks herself
In a few words, try to describe what I want to happen; what impact do I hopeto have? What do I want my reader to feel, or think, or maybe even do afterreading my paper? If my goal is to make an engineer, say someone like mycollege roommate, understand why people read poetry and maybe even come toenjoy it, how am I going to do this?
Clearly a review of the history of poetry won't do the trick. An Impact Statement
lets the writer plan with the reader in mind.
96. A Plan To Say
Once a writer has at least a start on what he or she wants To Do, i. makes sense
to try to plan what To Say. Here we are on the old familiar ground of outlines, but with
one distinction. Early in the writing process plans To Say might well be only sketches,
notes with arrows and stars. As models to be tested and changed and relined, they need
to be cheap enough to throw away. The later, more formal kinds of plans are really aids
to constructing tight, coherent text rather than generating possible ideas.
The following plan was done by a writer who liked to visualize his relationships.
Notice how it generates a set of things the writer could say in a personal profile, but
keeps those things within the context of what he wants to do by writing. Again compare
the result of this plan to the things he could have said had he chosen to simply describe
'My Job at Goodrich- .
Facts
Worker for Goodrich-
responsible for projects
developed 6 impeovenents
all engineering involved
Concepts
responsible
innovative
64
applied general theory
to real problems
developed cost effective
solution
61
,------;'
Worked for D'Alecy
successfully developed program
sharpened analytical skills
learned to shape complex
results to a model.
experienced
adaptable and
flexible
9.7. A Plan To Discover
For some writers the best plan is to begin by writing immediately. And this brings
us back to the question at the beginning of this paper. ho " can we preserve and in fact
foster the freedom to explore?
One way as by helping writers to build plans To Discover. When good writers "just
start to write" they are in fact calling on a rather sophisticated set of composing plans
This prr ^,edure should not be confused with undirected free-association sometimes
encouraged as "free writing They are not simply writing down what comes to mind
Instead, they are setting up conditions for discovery We could imagine them working
under a set of plans or private mental instructions such as these
iP
* don t try to be perfect it, just write and see where it leads
- don't worry about spelling, punctuation, etc
- follow an idea out altil it gets cold, then go on to a hot one
- don't worry about coherence and precise connections yet
* then after a period, go back. iN Ot to revise your text, but to see what you'veturned up
( 65
62
- what ideas look more promising, interesting
- how does this all fit together
- what implications, new ideas could I draw from this
In other words, when people write To Discover, they are working under a 'creative and
sophisticated plan which helps them handle the act of discovery better by consciously
fostering it. They are telling themselves how to carry out their own composing process.
It is not surprising that many students confuse this process with simply sitting
down and producing a paper from the top sentence down. it would look much the same
from the outside. But it is what's going on inside that makes all the difference. A writer
who has learned to plan has gained a degree of control over his or her Own writing
process
Teaching stuoents how they can make plans To Do, To Say, or To Discover can
offer them a genuinely useful skill. However, there is a difference between the complex
planning process writers really use, and the specific parts of it we isolate to teach. All
teaching techniques should probably carry a "product warning", mine would be this
techniques in a textbook look neat and orderly, but the process of writing isn't. Good
writers don't follow recipes or sit down and do planning exercises as they writs For
example, they may well be doing all three kinds of planning described here at the same
tome. What they do possess, though. are a set of options and powerful techniques, such
as planning, which they :.an use when they need them whether it is to help them to get
started. to get out of a block, or to Just carry on. Knowing such techniques lets writers
control their own writing process more and gives them the freedom to choose
alternatives as they write.
63
References
Baker,. S. The practical stylist. New York: Thomas Crowell, 1969.
Bitzer, L The rhetorical situation. Philosophy and Rhetoric, 1968, 1, 1-14.
Britton, J.. Burgess, T., Martin, M., McLeod, A., & Rosen, H. The development of writingabilities (11-18). London: Macmillan, 1975.
Flower, L Writer-based prose: A cognitive basis for problems in writip0. College English,September 1979 tat 41, 19-37.
/Flower, L. Writer based prose' A cognitive basisfor proble s in vifroting. Manuscript in
preparation, 1979 Ca
Flower, L, & Hayes, J.R. The dynamics of composing: Making plans and Jugglingconstraints. In L. Gregg & E. Steinberg (Eds.), Cognitive processes in writing: Aninterdisciplinary approach. Hillsdale. NJ: Lawrence Erlbaum Ass6ciates, 1980.
Flower, L., & Hayes. J R. A co nitive process theory of writing. College Co position andCommunication, 1981, 3 365-387.
\Flower, L, & Hayes, J R. ProbI6-s lying strategies and the writing proceis College
English, 1979(a]. 39. 44 1.
Flower, L. & Hayes. J R. Processbased evaluation of writing: Changin he performance,not the product. America Educational Researcloci ion Convention. SanFrancisco. April 1979 (13). ______-
Hayes. J R., & Flower. L Identifying the organization of writing processes In L Gregg &E Steinberg (Eds ), Cognitive processes in writing: An interdisciplinary approach.Hillsdale, NJ: Lawrence Erlbaum Associates. 1980 .
Larson. R L The rhetorical act of planning a piece of discourse Beaver CollegeConference on Evaluation of Writing. Glenside. PA, October 1978
Lowenthal. D. & Vfason, P Academics and their writing Times Literary Supplement. June1977, Vol 24
Lowes, J L The road to Xanadu Boston Houghton Mifflin. 1927..1.
O'Hare. F Sentence-combining. Improving student writing without formal grammarinstruction. Urbana, IL' National Council of Teachers of English. 1973
Rumelhart. D. Notes on a schema for stories In 0 Bobrow & A. Collins (Eds.).
Representation and understanding New York: Academic Press, 1975.
Simon, HA. & Chase, W.G Skill in chess American Scientist. 1973, 61, 394-403.
67
THE WRITER'S LONG TERM
MEMORY
Knowledge of Topic
Knowledge of Audience
Stored Wroftrig Mont
TASK ENVIRONMENT
WRITING ASSIGNIONT
Topic
Audience
Motivolinco Cues
TEXT
PRODUCED
so FAR
PLANNibIGp4.oX ORGANIZING
1...
m GOAL
SETTING
TRANSLATING REVIEWING
LIVE ADING1
1 EDITING]
I 1M O U I T O R
1
Figure 2.1 Structure of the Vriting Model.
GENERATING
1q1141014.0.04
Cv...PC HI 141 401.
PPM
he roiCC C Won III
AA 4.10.41 11,1,0111 41111
KCW mutt
SIX( tileiI
At lectio(1) t t [mini
. Cuhr.Cool ht 0Per PhOuf
It S
OJut.t rut uV*1
COAT IlOt
F:Cure 2.2 The structure of
the Ccnerating process.68
180
( s i t
1.0
t
lb( 1+111Y ONOLA IMO St.sACt
rOSss1tt ALS/UT TO PatviOtISI.T
I ImS1 rotvOuStY HOIED totiCS
41.1 iDesc ,aIED 1014C SultOtiOlnat
TO PlisatJef
10.4"Is.
St ArICH 'OnP titvIOuSty
NOM 101.1taSWIROF104NLIC
rottstottTOPIC
A. .4l)1 SC., INC( 1vat 10 ( Co.'7.PC$41ato.S f &IL
1.
Cr.:,etI Atr:A1 111 tAr4 I
I or 91 h7t.,:t
1----r7"
..cpc 0,.(0. 0.1.i
it.k.
CM=tat HISS
.Acro..1 IOU YRA l
O. .... . 41 . dt . a .11.. 1./titIPUN.4." . NOVI 41
I
Figure 2.5 The structure ofthe Reviewing process.
rel.'"" .44I:01 L..mil kijiix.; 44..I.A. rlsirair
RE VIEWING
bo rCx 1 Situcnt °f iiI
Cull St 4ALut
EDITING
WIT 1011 STANDARD LANGUAGE CONVENTIONS
l(furnt.1 to note)(spetteng fault)Illorni4oll(oiJimila ',wit)itsulin.11)(tupvtitousi of word).etc.
* fix Spelling fault)* fix grammar fault)* south for alternetive)
EDIT =OR ACCURACY OF MEANING
i(ionn4,1 or note)(wrong word)it:ntat or note)tamLnouous word)L ot
* fix word)* remove ambiguity)
EVALUATE FOR READER UNDERSTANDING
1(1k);11.11)(toltwoJI of technical word) *. find mom common word)
itt.n.nall(iiiissino context) * supply conte: ;L ig.
EVALUATE FOP. READER ACCEPTANCE
iitvimdlArn.ti.r..41 oticnt.ive to reader) * soften)
'0401143110one nconsisient) * make tosfofmlcki.
ig,.re 2.6 The structure of Ole Editing process
c'':."n- altcrnative 7.cees cf. etiitng.
1{C.,i.;,-iLd 1...9uo9e In SII.i -. edit
7 :',... nkt,. r. ',Oil In SI NI .... toV114.f Wei
3 -L, lac ,l t.c tir,i p'0.14C:GMS ( The =.0 vari from 4-.,our to viiit.r. Lev ri.) 1 171
7 ). jf...n - ovi.ote) tern.rate)
e 17s.),..01 , oro.,,,.:el -- oroannel
9 i..,,ot ii.,0,!...iv) ii..nst.nul
10 ;.,,,,,t I (, wits%) a.'. iCV,I
Ficure 2.7 '!:cnitor.
Configuration 1 (Depth first)3, ( New element from translate4. New element front organize5. 1 New element from generate6. I Not enuugh material
(goal = review))(goal translate))- (goal = organize))- (goal = generate))
Codiguretion 2 (Get rt down as you think of it, than review)3 1 New element from generate -. (goal = organize))4. 1 New eleulent horn °spoon -. (goal = translate))
5. I Not enough material (goal generate))
6. 1 Etough material (goal = review))
Cunloguration 3 (Perfect first draft)3. I Not enough material (goal .1 generate)]
4 t Enouto material, plan ; )1 complete - (0031 ° organize))
5 1 New clement from translate (goal = review))
6 1 Plan complete - (goal is translate))
Configuration 4 (Breadth first)3. I Not enough malarial -
m
{goal :=organize))4. I Enough atertal, plan not complete -,
-. (goal = translate))5. t Plan complete6. I Translation complete -, (goal .. review))
i,4ure 2.8 klternate configuration for the
monitor.
SEC7,',7,q; I
0 , 4: i: c "4-._ ... - '2C 4 C, '.: ::::%I.::": tt/--. t t rl, C..4 I.: Is. ec..tc ...sr.-- wow.; : w z. 1
: i 44 I , 4
721% SL:E 1 I I 1 I I I I I ; I I 1 1 1
L C i 0:
1 ec; I I *
4 I 4
1 I Iw
I 1 1 i
SECTION 2
160 200 220 240 260 250 '500 B20
I 1 1-'-'111111T=:111. 10 1I
...I
.. e...,-. - c..I I* / I 01
1 7,oOgi COD1 / 0
i 1H44 I I I I
I
. I I I
I
I
I
1 1 1 iOw' 1 4 4
I I 1 r I-1 ° r- 1
1
I iI 11 1 I 1 ° 1i l 1' ;I 1 1° I.
t1
420 440 =EC'
::._=1`-; 11" i 1 1 1 1 i 1 1 1 1 1
C Itl.:.%;21'.1 I 1 1 1 1 I
I
I I 1 I I
;:11..r.F.L.:TE . i1 I -1°-- 1 ° 1- -°:i ° 1 °°-' 1 "i " i l- 1
.:: 7 EI--,..14, OD4 4 I
01 6 I i° 1° I ee */ DO
I C.C.'t-03 Vt.::::.r1e'.1% ct.d Mixed
=inure 2.9 Processes attributed to se;-ents by.
Author 2.
71
P C:17): z. :SS 4E
w()::.E$
ES
PPESSUPt ON SCP.:AtIthG t+4074v ATION DI:St..111PTION
Or:PAC:01C." Cet;DSE rcOti OR Or COVRSr
70,73 P00; KS COL$R St
76 1713,M,260
Gehl EN1 vsts83
PMILOSOF>IIttL ISSu£
184
FREEDOA VS CONSTRAINT
94
GENERAL V.ORRY COuCERN stilft:D
160;17 AN: 4800 oBjEcmfts
97. 99 SC:ENCE 10%,113, 203
AHD
TECHNOLOOT
95,166APPLI- ENCODING
Cc:10N le 7 , 4 0,120
m00£ RN 129, i37
ANOUaOLS
20
Fiflute 2.10 The writer's plan and the protocol sertents in
which the ideas were generated.
BLS! Gw-4
,-I'
1 L Thit it April, 1g17..nd A end; st dtur,F a preurcol or. motivation.2. W: 01, um the istac is mo:it alio:. and On pronion of nuni papers. for3 mc. rrtolit Lion btu as Ca.ncsieNiellors is the aeader.it incisive and grades that4. ve so 1't borer pan tha: , ant Fades... Urn. they kind elfS compc1 mc. that's re111; sa hat rneuitation is. urn, kind of to impel or sun or6 t. rnomcnturn. (Paw )01..1 suppose frorrs the aeadcmie pressure of the p-ades,7 I'm no: sure 'A httilt!1 thirst pc:tonal utisfactior. is importer... but I'm not
,ure uhallaa trema from academic nest. tt, 4ne ;fades. ce %tetherI9. ttould sa; personal utssfacuon it a major issue Ok. urn. Oh.
14. Not only do I vet1-itf.;sion fro:7s mt trades, but 1 .lso Fes atid.raion sn turning in scrncinins
i6 that is fro: c..:.10; So. sf h.ppy ts hen I unit a Food paper. it :call; doc:-.1
17 matte: u he: Lind of ;rale 1 Fes ba:k on sr. if rm. It:7;1 wish i: So. um. urn.
1£ lel 'a tct Urn. uhat are eleI'm thsn',..in; of, y t;+s ^F so It:Et personal
19 ..orif.,:tte.-. hcluten resture stsd SAL ;uots. 1,41 I'm not mall; sure
:( It. Cl. It. him )( s:
ZS Urn. but
26 Cr the IC.sC .1 urts...; p.per the tars p1are is for thas Fradc,
r rt's r tr:a 7-rtr Iasi: %e rIxtedr.;be dr.ats r.o:
Figure 3,1 Segments of a proto:o3,
W, 07 IV 4.7 eNP PCESSUREG GRADESPS PrPlSOI:LL St 7:SFLCT'eN0 r t,,:.s ITY
1. e1\
20
4I lyL. C.
self
(..;
1%cure 3,2 7rP ''ricer's eevelo-in! knot..2edce
structure.
73
I. I: Thee is April. 1977. and Windy is s protocol e.n mouation.W Ok. um. the issue t rnoination and the problem of roust papers For
3. ono motiation hctc at Camerre31ellon is the acadrno: pressure and trades thatI sec intolsed. so I'd better put ohm down and packs 1m. die) land ofS. compel mc. that's really %sisal mouvorou IS. uM. kind of to impel 01 soar' 016. a momentum !Luse.} Ok. I suppose from the academic venom of the grades7 I m not sure whstheo-1 Mink petsonol satisfaction n tmpor.4...t but I'm not8. sine %Anita that awns Isom ilcadcmg grosuacs and trados, or uhetherI9. %so:mild sa; remorse! satislartion :s a memo ossue Ok. urn. Oh.
10 f What atc ;nu thinking?
II I'm 11)111 10 think Of the rum sentence so sun ssitis Urn. sna)bc something
12 like. personal sansfacion is the major morn aunt force in tae sensing of my13 Nowt and teflon 01. I m truing to ihm;, ofok, I %sant to sontebou rt it Into14 the ecioccnic ptcinci, 40%. trar, re.r;bc not so can 01 Not only do 1 fcrIS. satisfaction born my grotto. but I also get satisfaction in turning in something16 that is gaol qUalitr SO, of I'm hap'sy uhen i %sew a rood papas. it really doesn't
17 matter %shei kind of rt.& 1 fat back on it. tf I rn bow %stet it. So. um. um.IS Ire's re. Um, .'hat arc the -..1i1 010114' Of, tritir to tam personal
19. sAlIJCif011 tvci%tcn 3cedtinic ritcimoc and thc Etadci, bet I'm not really sloe20 Low to do .r. hos. to branch sr I'm troll) hasnt a hard mac rating started21 Well. ma; be I'll just sane a hunch of 'cleat doss's. and ma; be try "o onocrt them after
2? 1 finish 01 When I fecl4shat l'c unite's a hlril nuties.. and I put on paten2, Iftee, pipet nat. to he craded. %Own 1 submit it. the trade is non24. alssa necessar, for die teacher to hest the sonic Os. ohms kind of
?S. _1'11 citeet unh that ctic Os., atsclEel's sec %shit else. Um. but26. of courtc, the season I'm unroll the pert on the first place is for that vide.27 'nt to Matt that beet Thotc sue eic2c arc vcr) tot,IICKI.ed.(10)1C 0131% not28 the right Icon I'm. ot. I'm not al% a.s sure %sherbet nt) personal sausGarOn.
29 this n kind of off on a tantem, .id it rnotht not be included in IV) focal draft,
Figure 3.3 Wendy protocol.
74
InsNatural
Sentences from SAM Version:
2. Because of the emphasis on 4 Os here at ClAU, grades become an instinctivemotwatot for myself.
4. The initial motivator in the outset of writing a paper is the fact that agrade will be attached to it upon completion.6. Alter I begin writing a paper, the grade emphasis diminishes and ahigher level of personal satisfaction takes over.
Protocol Excerpt:
Urn because of tbe. maybe because of the empltasrsUrn 4 0's. Trying to bemore specrlic. Urn, even though I don't have a 4.0. Urn because of the emphasison 4.0*s. Ok, because of the emphases on 4 Ws, grades are maybe a natural, ormaybe instinctive, are instinctive motivator.
So urn, ok Th. 'mita/ motivator, thss is. grades are natutal instincsive rowel/mot.should say rnsirally again, but not really sure how to say it Ok. maybe I canleave it new-al, or instenrove--maybe that'll bring that Out. Urn, ok. rna0e I canrefer toward 'morally again Alter the urn. alter I begin porhaos, woding At% a
Paper, the fdcl that a grade, watt a minute. the fact that a grade. I don't knew if Iused later, attached-0k I did. I don't warn to use that again Ok. urn
After a begin writing a paper the glade ealphasisf don't know el I want to usethat main --the tootle emphasis is fOoeshado% eed by the by the fact or Maybel'fi come back and put that an, oh, Ok Alter I begin writing a paper the gradecrytphes.s is loreshaclowed by the something. urn, sorneitong that instead.1.11 have IV look up the vtading
vignrc 3.4 Turn[np tliourht.: int() Acceptable sentoncr.
""Phk f,"5",
75
J
The Rhetorical ProblemFitments of the Problem Examples
iTHE il...HriORICAL SITUATION]Exiger.cy 0t .A.ssignment
Audience
"Write for Seventeen tr.2..sar.ioe; thisis irnpossibk."
-Som:one like myself. but tditstedfor twenty years."
I
THE NI TER 5 OWN GOALS]invok in:. the
Reader "I'll chtne their notion of Engshteachers ..."
Persona or StV "I'll look Re 2.-) idiot if I say ..."
Hcz.n:ng So if I comp:re :hose two aTti-tudes . .."
1
Tral ' hrs: u 4-1: urn: tn Jr:rode:I:on
7ig...:re 4.1 Ele:-,ents f the rhetorical prob./a= writers:eon's'- to ther,selves in cc cosine.
, .A.:.1d)cr,ce S.::1 Text Mea-'-1.-
:;n 4'> s of rhe:o:ica! Art!ysis of Foals
2=...d .A. ssi z.:-.rnt :1; Tc:1.1s :.'.: i0 .A. -,4st nee
*,..1
0 0 3 7I
17
IE 1 11 1 5 ci 42
.1....z., zre .2. ::1..-1:,c..r cf t inES writer e:-7.1icf: tly -4.-.1-.resented
e.".) ?t-ect 0` .1-'P 1.....e:crf.cal 7:et...le in f!rst6,0 ilries cf protocol.
76
Pt,
TO DO
(RHETORICAL P1-414)
TO SAY To co14PDSE
IPPODVC7 -SLSEO) (PADCESS-eASED)
OENERLTE PRODUCEKRDwIED'..A XT Figure 5.1 Plan for dealing
with a Rhetorical Problem.
, TG op To cornzas TO SAY
h:t the SI.`,:V::::". totrkt r..,oloierri. atter-r.ale a;--oach is tosay
(
t .;-:t- Y. et ;-t el:
so these aPPlaetiorts aren't irri-:.s.r.ar..1 What is1-7.:-.am is vle
ahJe
c'C -;
5.' a 7rctccl'a: "crl'ing *,:y plan.
PLANS: 70 DO 70 COMPOSE 70 SAY
1, Wine an exposohonfor humanitiesteachers aboutMemory I fa groupof specializedmemory techni-ques).
2, What I'll do is jotdown randomthoughts about
3. What the teachersmight want to hearand I might want toto tell them.
4.
5. That word meansa lel but I won'texplain it now.
6.
7.
8.
9. One of the problemsin writing this essaywill be 10 expandon that usefulnessand mates' seemmore plausible. Tomake uses moregeneral and accept..able
7 6 Thal's the wrongword. s mean im-portant seeming
Uh Or. of that's
A iy)ont I will wantto rnake someplaceis *hal
Fact thong thatOccurs randomlyis encoding.
Thinking about ob-jections heard atthe workshop. Rolememory is trivial.
Memory I proce-dures are useful inmodern language.They are also moreuseful generally.
Unlonunately, bymore getteraty Imean thing% Megrocery II5
(I' Ontinue CII
PLANS: TO 00 TO COMPOSE TO SA
I. If we were to describeMemory I !as an example)what do we want 10 sayabout?
(Searches notes)lheoe it is. Leo's dothat.
2.
3. Let's see . nowwnai we Want 10 Ottacross in this plan.we want to Illustratethe practical nature
nature of skills
All right. / will taiketrakeas an example of Ilse sons ofmaterial . . , presented inthe course ... Now his isa Ierrible sentence but we canrevise it.
... the lint subunit of... unit... unit.
'That's not guile right, but, .
called Memo' y I In Memory I,Memo') :.. In Memory I thestudents learn ,_
Now wt, it am I going to dohere cause I don't Rally havean organization !co Memory Iyet.
... the students learn ...Now al this point we shouldbreak off and plan.
Note. Lines In nahcs arc Ir.orjrne.nis of the 0/Owing text interrupted by plans arid it-
comrnerds
1.110,ort 5.3 Prot ()col of the beginning of a w rl I. Inv:;en:, too.
Figure 6.1Episodes in an Expert Writer's Protocol.
Episode I My Job for a young - Oh I'm to describe my job for a
young, thirteen to fourteen year-,old teenage female
audience - Magazine - Seventeen. -a- My immediate
reaction ts that its utterly impossible. I did read
Seventeen, though - I guess I wouldn't say I read it -a-\
I looked at .t, especially the ads, so the idiawould be
to describe what I do to someone like myself when I read
- well not like myself, but adjusted for - well twenty
years late:. -a- Nov what I think of doing really is
that - until the coffee comes I feel I can't begin, so I
will snut. the door and feel the: I have a little bit more
////sode 2 TrIvaci, -1:7.:- Also the mention of a free-lance writer
is something I've - I've no experience in doing and my
sense .s that its a - a formula which I'm not sure I
know, so I Suppose what I have to do as -a- invent what
the formula might e, and - and then try to -a- try to
Include - events or occurzences or attitude or
ex:.--er.:ences In ry c,:n -,c'D tna._. would -a- that could ''..e -
//iiE7..!scfe 3a -.*.a: cc..:d te ;'era. .1n fc77..::a s^ la"e sea - I .
z.;:7-', -ne ...:-.:::: %an-. tc s-.art - ty w:I:lrs scretn:g -
t.7.a: Id%-: -a- a.._ a:. -.'-.e Q..= ....on of :r.e :eader - cf
t.'.al rCE7:f7 6.71 -a- : s::-.cse T.!%e. r.t.s: lnee7est:.ng t:17.
at..-. :ct: ....c,L::e 1-...s tnat :.: Is %g!..)y unlIkely tnat ...:
:::-::: s:t7 a: al: interes:1:; to sc-....".e of .hE: age - Fo
: chi zta:: t,:. say1:-.; scnetnIng :.,. ; :e - Can ycu Ira7:ne
s.:rFe:f szerdIrri Ya-y dv's 2.11 :e ::,:s - al:1;:c 1::.,
400BEST COPY Willa
E a7:. ht_ze. :eIlir; In yr.:r
cats... -a- walking out - out with coffee and a book and
watching the dawn materialize...I actually do
this...although 4:30's a bit early, perhaps I should say
5:30 so It won't seem - P.1-hough I do get up at 4:30 -a-
watching the dawn materialize and s_ar:inc to work - to
work by reading - reading the manuscript - of a Victorian
writer...with a manuscript of a...a Victorian writer...a
person with a manuscript of a student - Much like
yourself - Much like - Much like -a- a student or a book
tv ;:ristotle they've heard of Aristotle or - who could I
have .t be ty Plato.ozqatly vhen it gets to -Whon
you've... -a- fi.-.1cu-Idt coffee and whatever yoU had
to do (Oh thanks) - wnazever - now I've gotten just
coffee v-ur coffee (7u7.bling)...when you've
finiehed your coffee and -a- foreseen - and -a- ur,--mmmm -
when you've fsnithed your ocfee, you dress and drive -
ausou: three 7-_:es to the university w!ere szend
another - vh,re you spen! - spend !-ours - you spend
7:_rs a::-t tnrse a !my a-:v. three
-cre ni_rs :c ettdents - talking
istSi Wri
Episode 3b to - talking to other teachers...Um -/ should I (mumble)
- the thing is about saying teachers - the - the teenage
girl is going to think teachers like who she has, and
professor I always feel is sort of pretentious and a word
usually - usually I say teacher, but I know that means
I...It's unfortunate now in society we don't - but that
Episode 3c that isn't a prestige occupation./ Talking to other
people like yourselves - that's whoever it may be - other
people at your :cb - otner - other people like your sea; -
'.th a lot like yourseI' but - talking to other people like
yourself - going to meetings...coraittee meetings...and
doing all thir fc: nine months so that the other
ttree,..and doing all tnis for three months - okay - nine/non-its ...If you can inagine that....
Plan
1. The controversy about abortion
2. The con-T..i.tion
a. historically
1. religious.reason
2. economic reason
3. pro-14`e view
b. currently
1. religious
2. pro -life view
c. scaner line
Essay
1. An example of the controversy
2. The con-position
a. historically
1. religious reason
2. economic reason
3. pro-life view
3. Coum:er-argument 3. Ccumter'argument
a. Economic reason obsete'
b. religion less influential a religion less ialluential
c diversi:y cm moral issues b. diversity on moral issues
d. Supreme Cnur: view C. Supreme Court view
4. -we proposition L. The p:0-position
prcblems of unloved cnildren G. problems of unloved children
.:diems of poverty and ae b. problems of poverty
nc:hers
fc=t1c: --cw
c. proble=s of yo...1r:g and o::d
C. u=a7ried =ct^ers
e. fer
F Ccmclusionn;;;;cr
::::.;cr.ore of 7.1an and essay.
...The biggest mistake that beginning writers made is that they try to
write as -let's see - -write the first thing that comes into their mind--
write as soon as anything comes into their mind--try to write...that they
try ?o write...wri_e dog n whatever comesbut it's not that they write
down whatever cores into their minds- -some of them do thatyeahlike
uh--but some of the= are afraid to write anything dowm,..Okayso they
write a sentence at a -.-ime...The biggest mistake tba beginning writers
make is that they r...y to write without looking aheadyeah, I guess that's
better...they try to write without looking aheadthey onlythey or1y
think. cne sentence at a time and don't see where the next sentence is
going to lead than...
Figure 8.2 A In-ctocci seg=ent in which the writer appears to clarifyhis thinkins while ccnpcsing a sentence.
P:ctoccl:
Sf.t.tanco:
3 2 3
The c ast tn ing abcuz it is thatwhat? Sznathing abcu" using
4
allows me t:-.e zo--uh--I want tc write
S 6
Sc.matning er.put wy ideas--to ideas into action or to develop
7 E s :c
ideas intc-what? into a nAaninsful fcr=? Ch,
3 1 '.3 I+ ISt =a! urc--:.2.--al:cs =ascratch that.
:s
rte: Zttut i: is that it tc use/ n-y and
tdcas in a ---Vie. way.
The
:he cast it is
id Eat in a ;:.-scl,tti:e way.
al:cws me to use m-v mind and
exa--7.2a cf cent:nce gcncraticn.
START
.11
GET NEY.T ?LAN r1- ''`.. "7:ill no
no stored plan
t
1
52Rr'SS PI ^.14i
7 TV 7VT.... A. .
WRITE SEETENCE'
FROPCS:?ART
!ei lI
..INIM..
1
...111.=:....i
*-7.7---- ...7---=;D .0 C',:-.:..1 11
,... ... ... .......:.,.............
L
gios
,w11=11
rigure 8,4 ?:odel of Sentence Constluction.
-JO
TABLE 6.1Characteristics of Episodes
No. Mean il
Episodes Clauses/Epit.ode
DURATION OF EPISCDB8Range of (1:ime in minutes & seconds)
Clauses/ RangeEpisode
Expert
Vi25 9 1-23 22 sec. - 2 min.
8 min. 15 sec. 11 sec.
58 109 sec. -
1 - 4412 min.
1 min.
31 sec.
. , 7 sec. - 1 min.11 2 - 38 6 tin. 33 sec. 43 sec.
St,
Standa:dDeviation
2 mf.n.
S sec.N.
..
2 min.
1 sec.
1 .7.:::.
42 sec.
Oh, vhat can 1 say?--Drat, 1 broke the pencil point again - -keep
on breaking the pencil point - -I also have to-- i-e- drop the e- -to-do what.
is celled a - cuo:e hack-un - dash - seni-colonThis is a vav o: stcrint--. .. . .
-g- looks like an -f- on top - -o` stcrintuhthe,connuter--011, dratbrokk
4E again---for 4nCo.7ationuhto a roll of =aznetic tateMy -c- looks like
an -e- at the end cf zagnaric--I have to cnange thatLet ne get the eraser
out hereput it up on =7 pe%c42.--un--0%ayFeze ve goOkay, yhere a= 1 at?
Fiit.71 8.5 . 7.7cr atte=?tir4 to constr%:ct a sentence.
TABLE 2.1
Agreeme-at Among Raters in Assigning Properties to WrittenItems
/{V1 r Cala of $I W4 4.01
MarOtonws
A
935
1 A, 1 .93S
3 903
tt Law 146,-;
t 4th ott .924
"'Old droI
.9(4 CX&I
.910 952
9 S2
.v73 YA
TABLE 2.2Pro:ortion of Written Items With Each Property
.. 11 t,6,.. 4
..... W.I.
Sri 114J0
3S0 1540 CVO
Set own 2 Sri own J
004U90
b..*
0 6t70 LWD417
V.T,'E 2.3
7:c7.c:7?flOn. c Segr.enzs Assigned to Each P:ocess
1 2 St. ht.r1
1 4 1 A. 2 4. ,..t
.1:
L
: 1; 7 )
BEST COP1
TABLE 6.2Percent of Episode Boundaries Agreed Upon by Knowledgeable
Judges
WriterVo. of *
Jud esExpected
g% Agreement
Agreeing
Actual% Agreement
Expert 1 4
3
2
1
0
12
57
Expert 2 4
3
2
1
0
1
43
56
5A30 I
67%32
33
30
2070%
20
ao
Exile r' 3 4 0
3 1
2 12
1 b7
236
206t
34
Exert 4 4 0
3 1
2 10
1 69
Arcra:-c cf.
v---.--s
2 c.:
:-::-&.20
21
19 75%35
25
83
..-
Totalflew
ItIcaa
Reaponne to ) Reoponae toIthetorlent Reapoollo to Corraei
Problem Topic El clue it c
I Hete pont; c X Kt:upon:lcaht Lori co I to Top i c L
1.' r LA111:fit Current Clement
Cood
WrItera Cl 15H 106 48 67X 3374
62 9] 56 30 7 60 40
C3 115 66 45 57 42
Cl, 85 1.5 )6 4 5-.1 47
.11=1.11
Average 113 66 40 5 61X 119X
... .111... gr..... ........
Poo(Urigero P1 129 12 .66 41 17X 03X
PZ 05 20 6.1 2 24 16
P3 53 15 3I 1 20 12
PA 134 41 11 16 31 L9
P5 61 31 36 0 46 5411.11.
Avologe 94 26 56 1: MI 12Z
_Q)TABLE 7.1 Differences io How Good oi)d Poor WriLers GeneroLe Uew ideas
......... . ..,. .0.
ctoup
it.it Jo
?gni
J1
I
c.
TABLE 8.1Chunk Size and Essay Length of ExpArt and Novice Writers
11.
S Chunk
1 .
Ex-perts
c47e 1
..
zssav Lanzth S
Others
chl.mk Size 1 Essay Length
1 12.06 893 7 9.26 522
2 16.79 912 8 5.95;!
377.f.
3 13..c4 760MP
9 11.09 413 .
.
4 11.54 939 10 5.97,
451
5 8.68 656 11 7.10 704
6 4.76 553 12 .4.51 317
I.
;